{ "generatedAt": "2026-07-18T00:55:13.689Z", "source": { "name": "Epoch AI — Notable AI Models", "publisher": "Epoch AI", "url": "https://epoch.ai/data/ai-models", "downloadUrl": "https://epoch.ai/data/notable_ai_models.csv", "docsUrl": "https://epoch.ai/data/ai-models-documentation", "license": "CC BY 4.0", "accessed": "2026-07-18", "caveat": "Training-compute figures are Epoch AI estimates with varying confidence; not all models report compute/cost/power." }, "methodology": { "computeField": "Training compute (FLOP)", "doublingTimeMethod": "OLS regression of log10(training compute) on decimal year", "modernEraStart": 2010, "costUnit": "2023 USD", "notes": "Descriptive historical trends only — not a forecast.", "recentWindow": { "years": 3, "start": "2023-06-09", "end": "2026-06-09" } }, "definitions": { "frontierDefinition": "Epoch AI's 'Frontier model' flag marks models in the top 10 by estimated training compute at the time of release. It reflects compute-disclosure availability and historical compute scale — not capability, quality, or societal impact. frontierCount is derived from compute-known rows only; models without compute estimates (including many recent and open-source releases) cannot carry this flag.", "orgLeaderboardMetric": "modelCount = all tracked Epoch AI rows with a valid publication date, regardless of compute disclosure. computeKnownCount = rows with training compute estimates. frontierCount = compute-known rows flagged as top-10 compute at release. recentCount = full-catalog models published within the 3-year recent window.", "countryLeaderboardDefaultSort": "Default sort: recentCount (full-catalog models in the recent window) descending. This reflects current tracked-output activity. frontierCount is provided for historical context but must not be used as a general country ranking.", "openWeightsMetric": "openWeightsCount derives from Epoch AI's 'Open model weights?' column ('Yes' = confirmed open weights). A proxy for tracked open-release activity only — not downloads, adoption, quality, or societal impact.", "multiCountryAttribution": "Models attributed to multiple countries are counted once per participating country (co-attribution). A single US–UK collaboration increments both country totals.", "googleEntitiesNote": "Google, DeepMind, Google Brain, Google Research, and Google DeepMind are preserved as distinct Epoch AI source entities. No editorial merger is applied.", "coverageNote": "Coverage is Epoch AI's 'Notable AI Models' curation — not an exhaustive registry. Labs without compute disclosure and newer open-source releases may be underrepresented in the compute-known subset." }, "counts": { "totalRows": 1035, "withDate": 1030, "withCompute": 528, "withComputeAndDate": 528, "withPower": 215, "withCost": 179, "withOpenWeights": 318, "countries": 35, "recentWindowStart": "2023-06-09", "recentWindowEnd": "2026-06-09", "recentWindowCount": 317 }, "models": [ { "name": "Theseus", "organization": "Bell Laboratories", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "1950-07-02", "year": 1950, "decimalYear": 1950.49863, "domains": [ "Robotics" ], "task": "Maze solving", "parameters": 40, "computeFlop": 40, "log10Compute": 1.602, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.technologyreview.com/2018/12/19/138508/mighty-mouse/" }, { "name": "Perceptron Mark I", "organization": "Cornell Aeronautical Laboratory,Cornell University", "orgCategory": "Academia,Academia", "country": "United States", "countries": [ "United States" ], "date": "1957-01-01", "year": 1957, "decimalYear": 1957, "domains": [ "Other" ], "task": "Binary classification", "parameters": 1000, "computeFlop": 694894.9377361819, "log10Compute": 5.842, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://blogs.umass.edu/brain-wars/files/2016/03/rosenblatt-1957.pdf" }, { "name": "Pandemonium (morse)", "organization": "Massachusetts Institute of Technology (MIT)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "1959-02-01", "year": 1959, "decimalYear": 1959.084932, "domains": [ "Language" ], "task": "Morse translation", "parameters": null, "computeFlop": 600000000, "log10Compute": 8.778, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://aitopics.org/doc/classics:504E1BAC/" }, { "name": "Samuel Neural Checkers", "organization": "IBM", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "1959-07-01", "year": 1959, "decimalYear": 1959.49589, "domains": [ "Games" ], "task": "Checkers", "parameters": 16, "computeFlop": 428400000, "log10Compute": 8.632, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://ieeexplore.ieee.org/abstract/document/5392560" }, { "name": "Perceptron (1960)", "organization": "Cornell Aeronautical Laboratory", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "1960-03-30", "year": 1960, "decimalYear": 1960.243169, "domains": [ "Vision" ], "task": "Image classification", "parameters": 1000, "computeFlop": 720000000, "log10Compute": 8.857, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://www.semanticscholar.org/paper/Perceptron-Simulation-Experiments-Rosenblatt/ae76ce1ba27ac29addce4aab93b927e9bc7f7c67" }, { "name": "ADALINE", "organization": "Stanford University", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "1960-06-30", "year": 1960, "decimalYear": 1960.494536, "domains": [ "Vision" ], "task": "Pattern recognition", "parameters": 17, "computeFlop": 6600, "log10Compute": 3.82, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://isl.stanford.edu/~widrow/papers/c1960adaptiveswitching.pdf" }, { "name": "Linear Decision Functions", "organization": "Bell Laboratories", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "1962-06-01", "year": 1962, "decimalYear": 1962.413699, "domains": [ "Mathematics" ], "task": "Binary classification", "parameters": null, "computeFlop": 1559250, "log10Compute": 6.193, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://ieeexplore.ieee.org/document/4066882?denied=" }, { "name": "Print Recognition Logic", "organization": "IBM", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "1963-01-01", "year": 1963, "decimalYear": 1963, "domains": [ "Vision" ], "task": "Character recognition (OCR)", "parameters": null, "computeFlop": 22500000, "log10Compute": 7.352, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://ieeexplore.ieee.org/document/5392331" }, { "name": "Heuristic Reinforcement Learning", "organization": "Purdue University", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "1965-10-01", "year": 1965, "decimalYear": 1965.747945, "domains": [ "Robotics" ], "task": "System control", "parameters": null, "computeFlop": 1080000, "log10Compute": 6.033, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://ieeexplore.ieee.org/document/1098193" }, { "name": "LTE speaker verification system", "organization": "IBM", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "1966-11-01", "year": 1966, "decimalYear": 1966.832877, "domains": [ "Speech" ], "task": "Speech recognition (ASR)", "parameters": 2061, "computeFlop": 105917060, "log10Compute": 8.025, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://pubs.aip.org/asa/jasa/article-abstract/40/5/966/754180/Experimental-Studies-in-Speaker-Verification-Using?redirectedFrom=fulltext" }, { "name": "Cognitron", "organization": "Biological Cybernetics", "orgCategory": "Industry", "country": "Japan", "countries": [ "Japan" ], "date": "1975-09-01", "year": 1975, "decimalYear": 1975.665753, "domains": [ "Other" ], "task": "Miscellaneous image analysis,Image classification", "parameters": 21600, "computeFlop": 5184000, "log10Compute": 6.715, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://link.springer.com/article/10.1007%2FBF00342633" }, { "name": "Neocognitron", "organization": "NHK Broadcasting Science Research Laboratories", "orgCategory": "Industry", "country": "Japan", "countries": [ "Japan" ], "date": "1980-04-01", "year": 1980, "decimalYear": 1980.248634, "domains": [ "Vision" ], "task": "Character recognition (OCR)", "parameters": 1140576, "computeFlop": 273738240, "log10Compute": 8.437, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://link.springer.com/article/10.1007/BF00344251" }, { "name": "ASE+ACE", "organization": "University of Massachusetts Amherst", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "1983-09-01", "year": 1983, "decimalYear": 1983.665753, "domains": [ "Robotics" ], "task": "Pole balancing", "parameters": 324, "computeFlop": 324000000, "log10Compute": 8.511, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6313077" }, { "name": "Distributed representation NN", "organization": "Carnegie Mellon University (CMU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "1986-08-15", "year": 1986, "decimalYear": 1986.619178, "domains": [ "Other" ], "task": "Representation learning", "parameters": 432, "computeFlop": 388800000, "log10Compute": 8.59, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.cs.toronto.edu/~hinton/absps/families.pdf" }, { "name": "MLP with back-propagation", "organization": "University of California San Diego,Carnegie Mellon University (CMU)", "orgCategory": "Academia,Academia", "country": "United States", "countries": [ "United States" ], "date": "1986-10-01", "year": 1986, "decimalYear": 1986.747945, "domains": [ "Mathematics" ], "task": "Triplet completion", "parameters": 720, "computeFlop": 673920000, "log10Compute": 8.829, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Learning-representations-by-back-propagating-errors-Rumelhart-Hinton/052b1d8ce63b07fec3de9dbb583772d860b7c769" }, { "name": "NetTalk (transcription)", "organization": "Princeton University", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "1987-06-06", "year": 1987, "decimalYear": 1987.427397, "domains": [ "Speech" ], "task": "Speech synthesis", "parameters": 18629, "computeFlop": 28328002560, "log10Compute": 10.452, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "http://citeseerx.ist.psu.edu/viewdoc/download;jsessionid=03A3D3EDF0BAF35405ABCF083411B55E?doi=10.1.1.154.7012&rep=rep1&type=pdf" }, { "name": "NetTalk (dictionary)", "organization": "Princeton University", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "1987-06-06", "year": 1987, "decimalYear": 1987.427397, "domains": [ "Speech" ], "task": "Speech synthesis", "parameters": 18629, "computeFlop": 27664065000, "log10Compute": 10.442, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "http://citeseerx.ist.psu.edu/viewdoc/download;jsessionid=03A3D3EDF0BAF35405ABCF083411B55E?doi=10.1.1.154.7012&rep=rep1&type=pdf" }, { "name": "Translation-invariant MLP", "organization": "Carnegie Mellon University (CMU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "1987-06-15", "year": 1987, "decimalYear": 1987.452055, "domains": [], "task": "Object recognition", "parameters": 816, "computeFlop": 18032947200, "log10Compute": 10.256, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.cs.toronto.edu/~hinton/absps/parle.pdf" }, { "name": "MLN-ASR", "organization": "McGill University", "orgCategory": "Academia", "country": "Canada", "countries": [ "Canada" ], "date": "1988-08-01", "year": 1988, "decimalYear": 1988.581967, "domains": [ "Speech" ], "task": "Speech recognition (ASR)", "parameters": 10000, "computeFlop": 296425000, "log10Compute": 8.472, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://aaai.org/papers/00734-aaai88-130-data-driven-execution-of-multi-layered-networks-for-automatic-speech-recognition/" }, { "name": "Invariant image recognition", "organization": "Complutense University of Madrid", "orgCategory": "Academia", "country": "Spain", "countries": [ "Spain" ], "date": "1989-06-18", "year": 1989, "decimalYear": 1989.460274, "domains": [ "Vision" ], "task": "Representation learning", "parameters": null, "computeFlop": 27000000000, "log10Compute": 10.431, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://ieeexplore.ieee.org/document/118669" }, { "name": "Speaker-independent vowel classification", "organization": "University of Washington", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "1989-11-27", "year": 1989, "decimalYear": 1989.90411, "domains": [ "Speech" ], "task": "Speech recognition (ASR)", "parameters": 3040, "computeFlop": 7485696000, "log10Compute": 9.874, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Performance-Comparisons-Between-Backpropagation-and-Atlas-Cole/e42d2b89fcb4a1a3dfa63408f424f76975ed1e1b" }, { "name": "Handwritten digit recognition network", "organization": "AT&T", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "1989-11-27", "year": 1989, "decimalYear": 1989.90411, "domains": [ "Vision" ], "task": "Digit recognition", "parameters": 2578, "computeFlop": 181440000000, "log10Compute": 11.259, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Handwritten-Digit-Recognition-with-a-Network-LeCun-Boser/86ab4cae682fbd49c5a5bedb630e5a40fa7529f6" }, { "name": "Zip CNN", "organization": "AT&T,Bell Laboratories", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "1989-12-01", "year": 1989, "decimalYear": 1989.915068, "domains": [ "Vision" ], "task": "Character recognition (OCR)", "parameters": 9760, "computeFlop": 1496338054440, "log10Compute": 12.175, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://ieeexplore.ieee.org/document/6795724" }, { "name": "NETtalk reimplementation", "organization": "Oregon State University", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "1990-06-01", "year": 1990, "decimalYear": 1990.413699, "domains": [ "Speech" ], "task": "Text-to-speech (TTS)", "parameters": 27480, "computeFlop": 35811936000, "log10Compute": 10.554, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.sciencedirect.com/science/article/abs/pii/B9781558601413500079" }, { "name": "Bankruptcy-NN", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "1990-06-17", "year": 1990, "decimalYear": 1990.457534, "domains": [ "Other" ], "task": "Binary classification", "parameters": 36, "computeFlop": 3059337600, "log10Compute": 9.486, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/A-neural-network-model-for-bankruptcy-prediction-Odom-Sharda/ead9fa02902850a7418fb5ba720f3d9d8ab2f88b" }, { "name": "SexNet compression", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "1990-10-01", "year": 1990, "decimalYear": 1990.747945, "domains": [ "Vision" ], "task": "Image representation", "parameters": 72940, "computeFlop": 78775200000, "log10Compute": 10.896, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/SEXNET%3A-A-Neural-Network-Identifies-Sex-From-Human-Golomb-Lawrence/cbf90aa78fea0c8a1028705d92bc4bc7808ddeeb" }, { "name": "Weight Decay", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "1991-12-02", "year": 1991, "decimalYear": 1991.917808, "domains": [ "Speech" ], "task": "Speech synthesis", "parameters": 8386, "computeFlop": 75474000000, "log10Compute": 10.878, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/A-Simple-Weight-Decay-Can-Improve-Generalization-Krogh-Hertz/48e1de7d085808004d5f0493d486669a3d2930b5" }, { "name": "TD-Gammon", "organization": "IBM", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "1992-05-01", "year": 1992, "decimalYear": 1992.330601, "domains": [ "Games" ], "task": "Backgammon", "parameters": 25000, "computeFlop": 18232157622832.703, "log10Compute": 13.261, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://papers.nips.cc/paper/1991/file/68ce199ec2c5517597ce0a4d89620f55-Paper.pdf" }, { "name": "Cancer drug mechanism prediction", "organization": "National Cancer Institute", "orgCategory": "Government", "country": "United States", "countries": [ "United States" ], "date": "1992-10-16", "year": 1992, "decimalYear": 1992.789617, "domains": [ "Medicine" ], "task": "Drug discovery", "parameters": 594, "computeFlop": 53460000, "log10Compute": 7.728, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://pubmed.ncbi.nlm.nih.gov/1411538/" }, { "name": "Siamese-TDNN", "organization": "Bell Laboratories", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "1993-08-01", "year": 1993, "decimalYear": 1993.580822, "domains": [ "Vision" ], "task": "Image classification", "parameters": 744, "computeFlop": 12869570138112, "log10Compute": 13.11, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://www.semanticscholar.org/paper/Signature-Verification-Using-A-%22Siamese%22-Time-Delay-Bromley-Bentz/997dc5d9a058753f034422afe7bd0cc0b8ad808b" }, { "name": "ANN Eye Tracker", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "1993-11-29", "year": 1993, "decimalYear": 1993.909589, "domains": [ "Vision" ], "task": "Miscellaneous image analysis", "parameters": 5620, "computeFlop": 17534400000, "log10Compute": 10.244, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Non-Intrusive-Gaze-Tracking-Using-Artificial-Neural-Baluja-Pomerleau/574c0cf98825bf09b0ab7bbfe9ba89cd6090745e" }, { "name": "Ceramic-MLP", "organization": "Sapienza Università di Roma", "orgCategory": "Academia", "country": "Italy", "countries": [ "Italy" ], "date": "1994-01-07", "year": 1994, "decimalYear": 1994.016438, "domains": [ "Materials science" ], "task": "Pattern classification", "parameters": 1888, "computeFlop": 4531200000, "log10Compute": 9.656, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://www.sciencedirect.com/science/article/abs/pii/S0921883108605506" }, { "name": "Predictive Coding NN", "organization": "Technical University of Munich", "orgCategory": "Academia", "country": "Germany", "countries": [ "Germany" ], "date": "1994-12-02", "year": 1994, "decimalYear": 1994.917808, "domains": [ "Language" ], "task": "Language modeling", "parameters": 206910, "computeFlop": 18621900000000, "log10Compute": 13.27, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://proceedings.neurips.cc/paper/1994/file/5705e1164a8394aace6018e27d20d237-Paper.pdf" }, { "name": "NeuroChess", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "1994-12-02", "year": 1994, "decimalYear": 1994.917808, "domains": [ "Games" ], "task": "Chess", "parameters": 72251, "computeFlop": 858730812676, "log10Compute": 11.934, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://www.semanticscholar.org/paper/Learning-to-Play-the-Game-of-Chess-Thrun/4bc7a6dcb9e0e6c7a26800532e2a00f5572eea47" }, { "name": "Mixture of linear models", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "1994-12-02", "year": 1994, "decimalYear": 1994.917808, "domains": [ "Vision" ], "task": "Image classification", "parameters": 384000, "computeFlop": 453600000000, "log10Compute": 11.657, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://www.semanticscholar.org/paper/Recognizing-Handwritten-Digits-Using-Mixtures-of-Hinton-Revow/9dea20c1e5bbb1f543ff08113ffde5380c679f1f" }, { "name": "JPMAX", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "1994-12-02", "year": 1994, "decimalYear": 1994.917808, "domains": [ "Vision" ], "task": "Image representation", "parameters": 4446, "computeFlop": 80828280, "log10Compute": 7.908, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://proceedings.neurips.cc/paper_files/paper/1994/hash/4b0250793549726d5c1ea3906726ebfe-Abstract.html" }, { "name": "LISSOM", "organization": "University of Texas at Austin", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "1995-11-27", "year": 1995, "decimalYear": 1995.90411, "domains": [ "Vision" ], "task": "Digit recognition", "parameters": 432800, "computeFlop": 195544800000, "log10Compute": 11.291, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://www.semanticscholar.org/paper/Laterally-Interconnected-Self-Organizing-Maps-in-Choe-Sirosh/785f5facc76538ceba6f6b9e2d7b641d322e9854" }, { "name": "MUSIC perceptron", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "1996-06-03", "year": 1996, "decimalYear": 1996.420765, "domains": [ "Vision" ], "task": "Image completion", "parameters": 13607, "computeFlop": 881733600000, "log10Compute": 11.945, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://ieeexplore.ieee.org/document/549237" }, { "name": "System 11", "organization": "Carnegie Mellon University (CMU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "1996-06-18", "year": 1996, "decimalYear": 1996.461749, "domains": [ "Vision" ], "task": "Face detection", "parameters": 6452, "computeFlop": 25859616000, "log10Compute": 10.413, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://ieeexplore.ieee.org/document/655647" }, { "name": "SOM-CNN", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "1997-01-31", "year": 1997, "decimalYear": 1997.082192, "domains": [ "Vision" ], "task": "Face recognition", "parameters": 32015, "computeFlop": 31431600000, "log10Compute": 10.497, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Face-recognition%3A-a-convolutional-neural-network-Lawrence-Giles/86890c82b589e24007c56e1f40c5f928a0e04183" }, { "name": "LSTM", "organization": "Technical University of Munich", "orgCategory": "Academia", "country": "Germany", "countries": [ "Germany" ], "date": "1997-11-15", "year": 1997, "decimalYear": 1997.871233, "domains": [ "Language" ], "task": "Language modeling", "parameters": 10504, "computeFlop": 31512000000000, "log10Compute": 13.498, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://direct.mit.edu/neco/article-abstract/9/8/1735/6109/Long-Short-Term-Memory?redirectedFrom=fulltext" }, { "name": "LeNet-5", "organization": "AT&T", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "1998-11-01", "year": 1998, "decimalYear": 1998.832877, "domains": [ "Vision" ], "task": "Character recognition (OCR)", "parameters": 60000, "computeFlop": 2810937600000, "log10Compute": 12.449, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "http://vision.stanford.edu/cs598_spring07/papers/Lecun98.pdf" }, { "name": "RECONTRA-uncategorized", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "1999-06-02", "year": 1999, "decimalYear": 1999.416438, "domains": [ "Language" ], "task": "Translation", "parameters": 112000, "computeFlop": 3864000000000, "log10Compute": 12.587, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://www.semanticscholar.org/paper/Text-to-Text-Machine-Translation-Using-the-RECONTRA-Casta%C3%B1o-Casacuberta/47ee046b416d9258952cb8f4b0e2b6e65f334fad" }, { "name": "RECONTRA-categorized", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "1999-06-02", "year": 1999, "decimalYear": 1999.416438, "domains": [ "Language" ], "task": "Translation", "parameters": 66780, "computeFlop": 8013600000000, "log10Compute": 12.904, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://www.semanticscholar.org/paper/Text-to-Text-Machine-Translation-Using-the-RECONTRA-Casta%C3%B1o-Casacuberta/47ee046b416d9258952cb8f4b0e2b6e65f334fad" }, { "name": "PoE MNIST", "organization": "University College London (UCL)", "orgCategory": "Academia", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2000-11-28", "year": 2000, "decimalYear": 2000.907104, "domains": [ "Vision" ], "task": "Digit recognition", "parameters": 3925310, "computeFlop": 51810000000000, "log10Compute": 13.714, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://proceedings.neurips.cc/paper_files/paper/2000/file/1f1baa5b8edac74eb4eaa329f14a0361-Paper.pdf" }, { "name": "Neural LM", "organization": "University of Montreal / Université de Montréal", "orgCategory": "Academia", "country": "Canada", "countries": [ "Canada" ], "date": "2000-11-28", "year": 2000, "decimalYear": 2000.907104, "domains": [ "Language" ], "task": "Language modeling", "parameters": 6906980, "computeFlop": 6339000000000000, "log10Compute": 15.802, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://papers.nips.cc/paper_files/paper/2000/file/728f206c2a01bf572b5940d7d9a8fa4c-Paper.pdf" }, { "name": "Decision tree (classification)", "organization": "Mitsubishi Electric Research Labs,Compaq CRL", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2001-12-08", "year": 2001, "decimalYear": 2001.934247, "domains": [ "Vision" ], "task": "Face recognition", "parameters": 12000, "computeFlop": 63000000000000, "log10Compute": 13.799, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://www.cs.cmu.edu/~efros/courses/LBMV07/Papers/viola-cvpr-01.pdf" }, { "name": "NPLM (Brown)", "organization": "University of Montreal / Université de Montréal", "orgCategory": "Academia", "country": "Canada", "countries": [ "Canada" ], "date": "2003-03-15", "year": 2003, "decimalYear": 2003.2, "domains": [ "Language" ], "task": "Text autocompletion", "parameters": 4124233, "computeFlop": 132076260000000, "log10Compute": 14.121, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://dl.acm.org/doi/10.5555/944919.944966" }, { "name": "NPLM (AP News)", "organization": "University of Montreal / Université de Montréal", "orgCategory": "Academia", "country": "Canada", "countries": [ "Canada" ], "date": "2003-03-15", "year": 2003, "decimalYear": 2003.2, "domains": [ "Language" ], "task": "Text autocompletion", "parameters": 11904264, "computeFlop": 1666869200000000, "log10Compute": 15.222, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://dl.acm.org/doi/10.5555/944919.944966" }, { "name": "Invariant CNN", "organization": "New York University (NYU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2004-06-27", "year": 2004, "decimalYear": 2004.486339, "domains": [ "Vision" ], "task": "Object recognition", "parameters": 90575, "computeFlop": 974230000000, "log10Compute": 11.989, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Learning-methods-for-generic-object-recognition-to-LeCun-Huang/f354310098e09c1e1dc88758fca36767fd9d084d" }, { "name": "LMICA", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "2004-12-01", "year": 2004, "decimalYear": 2004.915301, "domains": [ "Vision" ], "task": "Object detection", "parameters": 4096000, "computeFlop": 2782080000000000, "log10Compute": 15.444, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Linear-Multilayer-Independent-Component-Analysis-Matsuda-Yamaguchi/7061b01572fbff2e223ce3abb59f397895b1ebf1" }, { "name": "Hierarchical LM", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "2005-01-06", "year": 2005, "decimalYear": 2005.013699, "domains": [ "Language" ], "task": "Language modeling", "parameters": null, "computeFlop": 115848000000000, "log10Compute": 14.064, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Hierarchical-Probabilistic-Neural-Network-Language-Morin-Bengio/c19fbefdeead6a4154a22a9c8551a18b1530033a" }, { "name": "RankNet", "organization": "Microsoft Research,Microsoft", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2005-08-07", "year": 2005, "decimalYear": 2005.59726, "domains": [ "Search" ], "task": "Search", "parameters": 5711, "computeFlop": 3482081588304, "log10Compute": 12.542, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://dl.acm.org/doi/abs/10.1145/1102351.1102363" }, { "name": "SVM-CNN", "organization": "New York University (NYU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2006-06-17", "year": 2006, "decimalYear": 2006.457534, "domains": [ "Vision" ], "task": "Image classification", "parameters": 90857, "computeFlop": 745200000000000, "log10Compute": 14.872, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Large-scale-Learning-with-SVM-and-Convolutional-for-Huang-LeCun/cf03fdf52dd6e4249cbbdbd0bffbbbe5ca389feb" }, { "name": "SB-LM", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2007-06-22", "year": 2007, "decimalYear": 2007.471233, "domains": [ "Language" ], "task": "Language modeling", "parameters": 300000000000, "computeFlop": 1449446400000000000, "log10Compute": 18.161, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://www.semanticscholar.org/paper/Large-Language-Models-in-Machine-Translation-Brants-Popat/ba786c46373892554b98df42df7af6f5da343c9d" }, { "name": "KN-LM", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2007-06-22", "year": 2007, "decimalYear": 2007.471233, "domains": [ "Language" ], "task": "Language modeling", "parameters": 21000000000, "computeFlop": 773038080000000000, "log10Compute": 17.888, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://www.semanticscholar.org/paper/Large-Language-Models-in-Machine-Translation-Brants-Popat/ba786c46373892554b98df42df7af6f5da343c9d" }, { "name": "GNN", "organization": "University of Siena", "orgCategory": "Academia", "country": "Italy", "countries": [ "Italy" ], "date": "2008-12-09", "year": 2008, "decimalYear": 2008.937158, "domains": [ "Other" ], "task": "Binary classification", "parameters": 30, "computeFlop": 1614600000, "log10Compute": 9.208, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://ieeexplore.ieee.org/document/4700287" }, { "name": "GPU DBNs", "organization": "Stanford University", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2009-06-15", "year": 2009, "decimalYear": 2009.452055, "domains": [ "Other" ], "task": "Miscellaneous image analysis", "parameters": 100000000, "computeFlop": 1000000000000000, "log10Compute": 15, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://dl.acm.org/doi/abs/10.1145/1553374.1553486?utm_campaign=The+Batch&utm_source=hs_email&utm_medium=email&_hsenc=p2ANqtz-95Z7-X4Dl-RJK6gYKvjyDIrYaGhBeqWoc0ldqyPEKni0Ip6UE7452hr-ygY52z00LBpYgM" }, { "name": "Two Stage Feature Extraction (MNIST)", "organization": "New York University (NYU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2009-09-01", "year": 2009, "decimalYear": 2009.665753, "domains": [ "Vision" ], "task": "Image classification", "parameters": 258800, "computeFlop": 20754000000000, "log10Compute": 13.317, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/What-is-the-best-multi-stage-architecture-for-Jarrett-Kavukcuoglu/1f88427d7aa8225e47f946ac41a0667d7b69ac52" }, { "name": "LCNP NORB", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "2009-11-22", "year": 2009, "decimalYear": 2009.890411, "domains": [ "Vision" ], "task": "Object recognition", "parameters": 16818176, "computeFlop": 2452090060800000, "log10Compute": 15.39, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://ieeexplore.ieee.org/document/5357786" }, { "name": "LCNP MNIST", "organization": "Unknown", "orgCategory": null, "country": null, "countries": [], "date": "2009-11-22", "year": 2009, "decimalYear": 2009.890411, "domains": [ "Vision" ], "task": "Object recognition", "parameters": 11616256, "computeFlop": 4181852160000000, "log10Compute": 15.621, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://ieeexplore.ieee.org/document/5357786" }, { "name": "LCNP LabelMe", "organization": "University of Bonn", "orgCategory": "Academia", "country": "Germany", "countries": [ "Germany" ], "date": "2009-11-22", "year": 2009, "decimalYear": 2009.890411, "domains": [ "Vision" ], "task": "Object recognition", "parameters": 13729792, "computeFlop": 3295150080000000, "log10Compute": 15.518, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://ieeexplore.ieee.org/document/5357786" }, { "name": "Feedforward NN", "organization": "University of Montreal / Université de Montréal", "orgCategory": "Academia", "country": "Canada", "countries": [ "Canada" ], "date": "2010-05-13", "year": 2010, "decimalYear": 2010.361644, "domains": [ "Vision" ], "task": "Digit recognition", "parameters": 7082000, "computeFlop": 350000000000000, "log10Compute": 14.544, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://proceedings.mlr.press/v9/glorot10a.html" }, { "name": "iCCCP", "organization": "Massachusetts Institute of Technology (MIT)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2010-06-13", "year": 2010, "decimalYear": 2010.446575, "domains": [ "Vision" ], "task": "Object detection", "parameters": null, "computeFlop": 1080000000000000, "log10Compute": 15.033, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://ieeexplore.ieee.org/document/5540096" }, { "name": "Pooling CNN (NORB)", "organization": "University of Bonn", "orgCategory": "Academia", "country": "Germany", "countries": [ "Germany" ], "date": "2010-09-15", "year": 2010, "decimalYear": 2010.70411, "domains": [ "Vision" ], "task": "Image classification", "parameters": 268664, "computeFlop": 1456277227200000, "log10Compute": 15.163, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Evaluation-of-Pooling-Operations-in-Convolutional-Scherer-M%C3%BCller/5d21006fa32ff69f6b0a646f26ce0db84f2f4d33" }, { "name": "Pooling CNN (Caltech 101)", "organization": "University of Bonn", "orgCategory": "Academia", "country": "Germany", "countries": [ "Germany" ], "date": "2010-09-15", "year": 2010, "decimalYear": 2010.70411, "domains": [ "Vision" ], "task": "Image classification", "parameters": 294912, "computeFlop": 1221124128768000, "log10Compute": 15.087, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Evaluation-of-Pooling-Operations-in-Convolutional-Scherer-M%C3%BCller/5d21006fa32ff69f6b0a646f26ce0db84f2f4d33" }, { "name": "RNN LM", "organization": "Johns Hopkins University", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2010-09-26", "year": 2010, "decimalYear": 2010.734247, "domains": [ "Language" ], "task": "Language modeling", "parameters": 70265000, "computeFlop": 53960000000000000, "log10Compute": 16.732, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://www.semanticscholar.org/paper/Recurrent-neural-network-based-language-model-Mikolov-Karafi%C3%A1t/9819b600a828a57e1cde047bbe710d3446b30da5" }, { "name": "Deep Autoencoders", "organization": "University of Toronto", "orgCategory": "Academia", "country": "Canada", "countries": [ "Canada" ], "date": "2011-04-29", "year": 2011, "decimalYear": 2011.323288, "domains": [ "Vision" ], "task": "Image representation", "parameters": 139808256, "computeFlop": 36728640000000000, "log10Compute": 16.565, "trainingCostUsd2023": null, "powerDrawW": 246.22598104391827, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Using-very-deep-autoencoders-for-content-based-Krizhevsky-Hinton/88080d28536f36588740737f3b7a1f6c1a409654" }, { "name": "High Performance CNN (NORB)", "organization": "IDSIA,SUPSI", "orgCategory": "Academia,Academia", "country": "Switzerland", "countries": [ "Switzerland" ], "date": "2011-07-16", "year": 2011, "decimalYear": 2011.536986, "domains": [ "Vision" ], "task": "Image classification", "parameters": 4878300, "computeFlop": 25798500000000000, "log10Compute": 16.412, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://people.idsia.ch/~juergen/ijcai2011.pdf" }, { "name": "CNN Committee (NIST)", "organization": "IDSIA", "orgCategory": "Academia", "country": "Switzerland", "countries": [ "Switzerland" ], "date": "2011-09-18", "year": 2011, "decimalYear": 2011.712329, "domains": [ "Vision" ], "task": "Image classification", "parameters": 128420, "computeFlop": 26000000000000000, "log10Compute": 16.415, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://www.semanticscholar.org/paper/Convolutional-Neural-Network-Committees-for-Ciresan-Meier/260a7615bbffec052d67dffde5bcf9b4687b50ee" }, { "name": "CNN Committee (MNIST)", "organization": "IDSIA", "orgCategory": "Academia", "country": "Switzerland", "countries": [ "Switzerland" ], "date": "2011-09-18", "year": 2011, "decimalYear": 2011.712329, "domains": [ "Vision" ], "task": "Image classification", "parameters": 120620, "computeFlop": 52000000000000000, "log10Compute": 16.716, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://www.semanticscholar.org/paper/Convolutional-Neural-Network-Committees-for-Ciresan-Meier/260a7615bbffec052d67dffde5bcf9b4687b50ee" }, { "name": "CNN committee (traffic sign)", "organization": "IDSIA", "orgCategory": "Academia", "country": "Switzerland", "countries": [ "Switzerland" ], "date": "2011-10-03", "year": 2011, "decimalYear": 2011.753425, "domains": [ "Vision" ], "task": "Image classification", "parameters": 1388800, "computeFlop": 991981425600000, "log10Compute": 14.997, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://www.semanticscholar.org/paper/A-committee-of-neural-networks-for-traffic-sign-Ciresan-Meier/dd7f8b53e6802787179a961e766760cbbe2d5011" }, { "name": "Dropout (MNIST)", "organization": "University of Toronto", "orgCategory": "Academia", "country": "Canada", "countries": [ "Canada" ], "date": "2012-06-03", "year": 2012, "decimalYear": 2012.420765, "domains": [ "Vision" ], "task": "Character recognition (OCR)", "parameters": 5594010, "computeFlop": 6039370800000000, "log10Compute": 15.781, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1207.0580" }, { "name": "Dropout (ImageNet)", "organization": "University of Toronto", "orgCategory": "Academia", "country": "Canada", "countries": [ "Canada" ], "date": "2012-06-03", "year": 2012, "decimalYear": 2012.420765, "domains": [ "Vision" ], "task": "Image classification", "parameters": null, "computeFlop": 273196800000000000, "log10Compute": 17.436, "trainingCostUsd2023": 7.952555832759744, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1207.0580" }, { "name": "Dropout (CIFAR)", "organization": "University of Toronto", "orgCategory": "Academia", "country": "Canada", "countries": [ "Canada" ], "date": "2012-06-03", "year": 2012, "decimalYear": 2012.420765, "domains": [ "Vision" ], "task": "Character recognition (OCR)", "parameters": null, "computeFlop": 4268700000000000, "log10Compute": 15.63, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1207.0580" }, { "name": "Unsupervised High-level Feature Learner", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2012-07-12", "year": 2012, "decimalYear": 2012.527322, "domains": [ "Vision" ], "task": "Image classification", "parameters": 1000000000, "computeFlop": 600000000000000000, "log10Compute": 17.778, "trainingCostUsd2023": 16.029257461780738, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://arxiv.org/abs/1112.6209" }, { "name": "LSTM LM", "organization": "RWTH Aachen University", "orgCategory": "Academia", "country": "Germany", "countries": [ "Germany" ], "date": "2012-09-09", "year": 2012, "decimalYear": 2012.688525, "domains": [ "Language" ], "task": "Language modeling", "parameters": 102720000, "computeFlop": 16600000000000000, "log10Compute": 16.22, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://www.semanticscholar.org/paper/LSTM-Neural-Networks-for-Language-Modeling-Sundermeyer-Schl%C3%BCter/f9a1b3850dfd837793743565a8af95973d395a4e" }, { "name": "AlexNet", "organization": "University of Toronto", "orgCategory": "Academia", "country": "Canada", "countries": [ "Canada" ], "date": "2012-09-30", "year": 2012, "decimalYear": 2012.745902, "domains": [ "Vision" ], "task": "Image classification", "parameters": 60000000, "computeFlop": 470000000000000000, "log10Compute": 17.672, "trainingCostUsd2023": 15.56975299843643, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://proceedings.neurips.cc/paper/2012/hash/c399862d3b9d6b76c8436e924a68c45b-Abstract.html" }, { "name": "DNN EM segmentation", "organization": "IDSIA,SUPSI", "orgCategory": "Academia,Academia", "country": "Switzerland", "countries": [ "Switzerland" ], "date": "2012-12-03", "year": 2012, "decimalYear": 2012.920765, "domains": [ "Vision" ], "task": "Image segmentation", "parameters": 218896, "computeFlop": 478000000000000000, "log10Compute": 17.679, "trainingCostUsd2023": 3.9045853984413377, "powerDrawW": 2116.2977934480996, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Deep-Neural-Networks-Segment-Neuronal-Membranes-in-Ciresan-Giusti/09193e19b59fc8f05bee9d6efbfb1607ca5b6501" }, { "name": "DistBelief Speech", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2012-12-03", "year": 2012, "decimalYear": 2012.920765, "domains": [ "Speech" ], "task": "Speech recognition (ASR)", "parameters": 47185920, "computeFlop": 311400000000000000, "log10Compute": 17.493, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://www.semanticscholar.org/paper/Large-Scale-Distributed-Deep-Networks-Dean-Corrado/3127190433230b3dc1abd0680bb58dced4bcd90e" }, { "name": "DistBelief NNLM", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2013-01-16", "year": 2013, "decimalYear": 2013.041096, "domains": [ "Language" ], "task": "Semantic embedding", "parameters": null, "computeFlop": 2612736000000000000, "log10Compute": 18.417, "trainingCostUsd2023": 3255.1839637787602, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://arxiv.org/abs/1301.3781" }, { "name": "ReLU-Speech", "organization": "Google,University of Toronto,New York University (NYU)", "orgCategory": "Industry,Academia,Academia", "country": "United States, Canada", "countries": [ "United States", "Canada" ], "date": "2013-05-26", "year": 2013, "decimalYear": 2013.39726, "domains": [ "Speech" ], "task": "Speech recognition (ASR)", "parameters": 101706240, "computeFlop": 127733760000000000, "log10Compute": 17.106, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://www.semanticscholar.org/paper/On-rectified-linear-units-for-speech-processing-Zeiler-Ranzato/64da1980714cfc130632c5b92b9d98c2f6763de6" }, { "name": "Hierarchical Scene Labeling (Stanford Background)", "organization": "New York University (NYU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2013-08-01", "year": 2013, "decimalYear": 2013.580822, "domains": [ "Vision" ], "task": "Semantic segmentation", "parameters": 51609600, "computeFlop": 237746880000000000, "log10Compute": 17.376, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://ieeexplore.ieee.org/document/6338939" }, { "name": "RNTN", "organization": "Stanford University", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2013-10-01", "year": 2013, "decimalYear": 2013.747945, "domains": [ "Language" ], "task": "Sentiment classification", "parameters": null, "computeFlop": 14220000000000000, "log10Compute": 16.153, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://www.semanticscholar.org/paper/Recursive-Deep-Models-for-Semantic-Compositionality-Socher-Perelygin/687bac2d3320083eb4530bf18bb8f8f721477600" }, { "name": "RCTM", "organization": "University of Oxford", "orgCategory": "Academia", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2013-10-01", "year": 2013, "decimalYear": 2013.747945, "domains": [ "Language" ], "task": "Translation", "parameters": null, "computeFlop": 9331200000000000, "log10Compute": 15.97, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://www.semanticscholar.org/paper/Recurrent-Continuous-Translation-Models-Kalchbrenner-Blunsom/944a1cfd79dbfb6fef460360a0765ba790f4027a" }, { "name": "Word2Vec (large)", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2013-10-16", "year": 2013, "decimalYear": 2013.789041, "domains": [ "Language" ], "task": "Semantic embedding", "parameters": 692000000, "computeFlop": 38880000000000000, "log10Compute": 16.59, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://arxiv.org/abs/1310.4546" }, { "name": "Visualizing CNNs", "organization": "New York University (NYU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2013-11-12", "year": 2013, "decimalYear": 2013.863014, "domains": [ "Vision" ], "task": "Image classification", "parameters": null, "computeFlop": 532000000000000000, "log10Compute": 17.726, "trainingCostUsd2023": 12.994002368456531, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1311.2901" }, { "name": "TransE", "organization": "Universite de Technologie de Compiègne – CNRS,Google", "orgCategory": "Academia,Industry", "country": "France, United States", "countries": [ "France", "United States" ], "date": "2013-12-05", "year": 2013, "decimalYear": 2013.926027, "domains": [ "Language" ], "task": "Entity embedding", "parameters": 942000000, "computeFlop": 1340928000000000000, "log10Compute": 18.127, "trainingCostUsd2023": 30.028097313317172, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://papers.nips.cc/paper/2013/hash/1cecc7a77928ca8133fa24680a88d2f9-Abstract.html" }, { "name": "DQN", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2013-12-19", "year": 2013, "decimalYear": 2013.964384, "domains": [ "Games" ], "task": "Atari", "parameters": 836096, "computeFlop": 2846883840000000, "log10Compute": 15.454, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1312.5602" }, { "name": "Image generation", "organization": "University of Amsterdam", "orgCategory": "Academia", "country": "Netherlands", "countries": [ "Netherlands" ], "date": "2013-12-20", "year": 2013, "decimalYear": 2013.967123, "domains": [ "Vision" ], "task": "Image clustering", "parameters": 784000, "computeFlop": 475200000000000, "log10Compute": 14.677, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1312.6114" }, { "name": "GANs", "organization": "University of Montreal / Université de Montréal", "orgCategory": "Academia", "country": "Canada", "countries": [ "Canada" ], "date": "2014-06-10", "year": 2014, "decimalYear": 2014.438356, "domains": [ "Image generation" ], "task": "Image generation", "parameters": null, "computeFlop": 518400000000000000, "log10Compute": 17.715, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://arxiv.org/abs/1406.2661" }, { "name": "SPPNet", "organization": "Microsoft,Xi’an Jiaotong University,University of Science and Technology of China (USTC)", "orgCategory": "Industry,Academia,Academia", "country": "United States, China", "countries": [ "United States", "China" ], "date": "2014-06-18", "year": 2014, "decimalYear": 2014.460274, "domains": [ "Vision" ], "task": "Image classification", "parameters": null, "computeFlop": 3411072000000000000, "log10Compute": 18.533, "trainingCostUsd2023": 51.65209223027738, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1406.4729" }, { "name": "SmooCT", "organization": "University College London (UCL)", "orgCategory": "Academia", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2014-07-01", "year": 2014, "decimalYear": 2014.49589, "domains": [ "Games" ], "task": "Poker", "parameters": null, "computeFlop": 69000000000000000, "log10Compute": 16.839, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://www.semanticscholar.org/paper/Self-play-Monte-Carlo-tree-search-in-computer-poker-Heinrich-Silver/7b687599b4425aa959036071030e1212a3b359c7" }, { "name": "ACF-WIDER", "organization": "Chinese Academy of Sciences", "orgCategory": "Academia", "country": "China", "countries": [ "China" ], "date": "2014-07-15", "year": 2014, "decimalYear": 2014.534247, "domains": [ "Vision" ], "task": "Face detection", "parameters": 6144, "computeFlop": 76380000000000, "log10Compute": 13.883, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1407.4023" }, { "name": "RNNsearch-50*", "organization": "Jacobs University Bremen,University of Montreal / Université de Montréal", "orgCategory": "Academia,Academia", "country": "Germany, Canada", "countries": [ "Germany", "Canada" ], "date": "2014-09-01", "year": 2014, "decimalYear": 2014.665753, "domains": [ "Language" ], "task": "Translation", "parameters": null, "computeFlop": 1555200000000000000, "log10Compute": 18.192, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1409.0473" }, { "name": "VGG19", "organization": "University of Oxford", "orgCategory": "Academia", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2014-09-04", "year": 2014, "decimalYear": 2014.673973, "domains": [ "Vision" ], "task": "Image classification", "parameters": 144000000, "computeFlop": 11000000000000000000, "log10Compute": 19.041, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://arxiv.org/abs/1409.1556" }, { "name": "VGG16", "organization": "University of Oxford", "orgCategory": "Academia", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2014-09-04", "year": 2014, "decimalYear": 2014.673973, "domains": [ "Vision" ], "task": "Image classification", "parameters": 138000000, "computeFlop": 12291000000000000000, "log10Compute": 19.09, "trainingCostUsd2023": 239.2376655010467, "powerDrawW": 2137.653074766455, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1409.1556" }, { "name": "Seq2Seq LSTM", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2014-09-10", "year": 2014, "decimalYear": 2014.690411, "domains": [ "Language" ], "task": "Translation", "parameters": 1920000000, "computeFlop": 56000000000000000000, "log10Compute": 19.748, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1409.3215" }, { "name": "SPN-4+KN5", "organization": "Singapore University of Technology & Design,DSO National Laboratories", "orgCategory": "Academia,Government", "country": "Singapore", "countries": [ "Singapore" ], "date": "2014-09-14", "year": 2014, "decimalYear": 2014.70137, "domains": [ "Language" ], "task": "Language modeling", "parameters": 5000000, "computeFlop": 44000000000000000, "log10Compute": 16.643, "trainingCostUsd2023": null, "powerDrawW": 290.04546165358084, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://www.comp.nus.edu.sg/~skok/papers/is14.pdf" }, { "name": "GoogLeNet / InceptionV1", "organization": "Google,University of Michigan,University of North Carolina", "orgCategory": "Industry,Academia,Academia", "country": "United States", "countries": [ "United States" ], "date": "2014-09-17", "year": 2014, "decimalYear": 2014.709589, "domains": [ "Vision" ], "task": "Image classification", "parameters": 6797700, "computeFlop": 1510000000000000000, "log10Compute": 18.179, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1409.4842" }, { "name": "TA-CNN", "organization": "Chinese University of Hong Kong (CUHK)", "orgCategory": "Academia", "country": "Hong Kong", "countries": [ "Hong Kong" ], "date": "2014-11-29", "year": 2014, "decimalYear": 2014.909589, "domains": [ "Vision" ], "task": "Object detection", "parameters": 706048, "computeFlop": 10854000000000000, "log10Compute": 16.036, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1412.0069" }, { "name": "SNM-skip", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2014-12-03", "year": 2014, "decimalYear": 2014.920548, "domains": [ "Language" ], "task": "Language modeling", "parameters": 62000000000, "computeFlop": 297600000001000000000, "log10Compute": 20.474, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://arxiv.org/abs/1412.1454" }, { "name": "Fractional Max-Pooling", "organization": "University of Warwick", "orgCategory": "Academia", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2014-12-18", "year": 2014, "decimalYear": 2014.961644, "domains": [ "Vision" ], "task": "Image classification", "parameters": 27000000, "computeFlop": 100000000000000000, "log10Compute": 17, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://arxiv.org/abs/1412.6071v4" }, { "name": "ADAM (CIFAR-10)", "organization": "University of Amsterdam,OpenAI,University of Toronto", "orgCategory": "Academia,Industry,Academia", "country": "Netherlands, United States, Canada", "countries": [ "Netherlands", "United States", "Canada" ], "date": "2014-12-22", "year": 2014, "decimalYear": 2014.972603, "domains": [ "Vision" ], "task": "Image classification", "parameters": 2370000, "computeFlop": 624979800000000, "log10Compute": 14.796, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1412.6980" }, { "name": "MSRA (C, PReLU)", "organization": "Microsoft Research", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2015-02-06", "year": 2015, "decimalYear": 2015.09863, "domains": [ "Vision" ], "task": "Image classification", "parameters": 87048800, "computeFlop": 23974030080000000000, "log10Compute": 19.38, "trainingCostUsd2023": 1394.115708804708, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1502.01852" }, { "name": "genCNN + dyn eval", "organization": "Chinese Academy of Sciences,Huawei Noah's Ark Lab,Dublin City University", "orgCategory": "Academia,Industry,Academia", "country": "China, Ireland", "countries": [ "China", "Ireland" ], "date": "2015-03-17", "year": 2015, "decimalYear": 2015.205479, "domains": [ "Language" ], "task": "Language modeling", "parameters": 8000000, "computeFlop": 34153451000000000, "log10Compute": 16.533, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://aclanthology.org/P15-1151/" }, { "name": "TC-DNN-BLSTM-DNN", "organization": "Carnegie Mellon University (CMU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2015-04-06", "year": 2015, "decimalYear": 2015.260274, "domains": [ "Speech" ], "task": "Speech recognition (ASR),Speech-to-text", "parameters": 18413568, "computeFlop": 194101919999999970, "log10Compute": 17.288, "trainingCostUsd2023": null, "powerDrawW": 263.01387537165505, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Deep-Recurrent-Neural-Networks-for-Acoustic-Chan-Lane/ae18f67c10a5d4da91f128ec7a6cf7c784122cd5" }, { "name": "U-Net", "organization": "University of Freiburg", "orgCategory": "Academia", "country": "Germany", "countries": [ "Germany" ], "date": "2015-05-18", "year": 2015, "decimalYear": 2015.375342, "domains": [ "Vision" ], "task": "Image segmentation", "parameters": 37676160, "computeFlop": 50832252000000000, "log10Compute": 16.706, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/U-Net%3A-Convolutional-Networks-for-Biomedical-Image-Ronneberger-Fischer/6364fdaa0a0eccd823a779fcdd489173f938e91a" }, { "name": "DCNN", "organization": "University of Maryland,Rutgers University", "orgCategory": "Academia,Academia", "country": "United States", "countries": [ "United States" ], "date": "2015-08-07", "year": 2015, "decimalYear": 2015.59726, "domains": [ "Vision" ], "task": "Face verification", "parameters": 5006000, "computeFlop": 480986910000000000, "log10Compute": 17.682, "trainingCostUsd2023": null, "powerDrawW": 285.60949180561556, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1508.01722" }, { "name": "AlphaGo Fan", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2015-10-01", "year": 2015, "decimalYear": 2015.747945, "domains": [ "Games" ], "task": "Go", "parameters": 8209984, "computeFlop": 380000000000000000000, "log10Compute": 20.58, "trainingCostUsd2023": 4828.322276803044, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://www.nature.com/articles/nature16961" }, { "name": "SAF R-CNN", "organization": "Beijing Institute of Technology,Sun Yat-sen University,Panasonic R&D,National University of Singapore", "orgCategory": "Academia,Academia,Industry,Academia", "country": "China, Singapore", "countries": [ "China", "Singapore" ], "date": "2015-10-28", "year": 2015, "decimalYear": 2015.821918, "domains": [ "Vision" ], "task": "Object detection", "parameters": 138000000, "computeFlop": 12311081250000001000, "log10Compute": 19.09, "trainingCostUsd2023": null, "powerDrawW": 290.9065505304121, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1510.08160" }, { "name": "Inception v3", "organization": "Google,University College London (UCL)", "orgCategory": "Industry,Academia", "country": "United States, United Kingdom", "countries": [ "United States", "United Kingdom" ], "date": "2015-12-02", "year": 2015, "decimalYear": 2015.917808, "domains": [ "Vision" ], "task": "Image classification", "parameters": 23626728, "computeFlop": 100000000000000000000, "log10Compute": 20, "trainingCostUsd2023": 1218.1810104797394, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://arxiv.org/abs/1512.00567" }, { "name": "ResNet-152 (ImageNet)", "organization": "Microsoft", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2015-12-10", "year": 2015, "decimalYear": 2015.939726, "domains": [ "Vision" ], "task": "Image classification", "parameters": 60200000, "computeFlop": 10414080000000000000, "log10Compute": 19.018, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1512.03385" }, { "name": "ResNet-101 (ImageNet)", "organization": "Microsoft", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2015-12-10", "year": 2015, "decimalYear": 2015.939726, "domains": [ "Vision" ], "task": "Image classification", "parameters": 44500000, "computeFlop": 7004000000000000000, "log10Compute": 18.845, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/1512.03385" }, { "name": "Variational (untied weights, MC) LSTM (Large)", "organization": "University of Cambridge", "orgCategory": "Academia", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2015-12-16", "year": 2015, "decimalYear": 2015.956164, "domains": [ "Language" ], "task": "Language modeling", "parameters": 66000000, "computeFlop": 5886144000000000, "log10Compute": 15.77, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1512.05287" }, { "name": "AlphaGo Lee", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2016-01-27", "year": 2016, "decimalYear": 2016.071038, "domains": [ "Games" ], "task": "Go", "parameters": null, "computeFlop": 1.9e+21, "log10Compute": 21.279, "trainingCostUsd2023": 22206.80954117433, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://www.nature.com/articles/nature16961" }, { "name": "Named Entity Recognition model", "organization": "Carnegie Mellon University (CMU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2016-03-04", "year": 2016, "decimalYear": 2016.172131, "domains": [ "Language" ], "task": "Named entity recognition (NER),Language modeling", "parameters": null, "computeFlop": 96940800000000000, "log10Compute": 16.987, "trainingCostUsd2023": null, "powerDrawW": 290.0785080846343, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1603.01354" }, { "name": "R-FCN", "organization": "Tsinghua University,Microsoft Research", "orgCategory": "Academia,Industry", "country": "China, United States", "countries": [ "China", "United States" ], "date": "2016-06-21", "year": 2016, "decimalYear": 2016.469945, "domains": [ "Vision" ], "task": "Object detection", "parameters": null, "computeFlop": 719357760000000000, "log10Compute": 17.857, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1605.06409" }, { "name": "ResNet-200", "organization": "Microsoft Research Asia", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2016-09-17", "year": 2016, "decimalYear": 2016.710383, "domains": [ "Vision" ], "task": "Image classification", "parameters": null, "computeFlop": 29741645000000000000, "log10Compute": 19.473, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://link.springer.com/chapter/10.1007/978-3-319-46493-0_38" }, { "name": "Pointer Sentinel-LSTM (medium)", "organization": "MetaMind Inc,Salesforce", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2016-09-26", "year": 2016, "decimalYear": 2016.734973, "domains": [ "Language" ], "task": "Language modeling", "parameters": 21000000, "computeFlop": 7490000000000000, "log10Compute": 15.874, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1609.07843" }, { "name": "GNMT", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2016-09-26", "year": 2016, "decimalYear": 2016.734973, "domains": [ "Language" ], "task": "Translation", "parameters": 278000000, "computeFlop": 6.620000000001e+21, "log10Compute": 21.821, "trainingCostUsd2023": 201331.569937687, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "Hosted access (no API)", "confidence": "Likely", "link": "https://arxiv.org/abs/1609.08144" }, { "name": "Xception", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2016-10-07", "year": 2016, "decimalYear": 2016.765027, "domains": [ "Vision" ], "task": "Image classification", "parameters": 22855952, "computeFlop": 436000000000000000000, "log10Compute": 20.639, "trainingCostUsd2023": 13108.852355728606, "powerDrawW": 37828.64014138488, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1610.02357" }, { "name": "SPIDER2", "organization": "Griffith University,University of Iowa,Dezhou University", "orgCategory": "Academia,Academia,Academia", "country": "Australia, United States, China", "countries": [ "Australia", "United States", "China" ], "date": "2016-10-28", "year": 2016, "decimalYear": 2016.822404, "domains": [ "Biology" ], "task": "Protein folding prediction,Proteins", "parameters": 409536, "computeFlop": 18220000000000000, "log10Compute": 16.261, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Likely", "link": "https://link.springer.com/protocol/10.1007/978-1-4939-6406-2_6" }, { "name": "VD-LSTM+REAL Large", "organization": "Salesforce Research,Stanford University", "orgCategory": "Industry,Academia", "country": "United States", "countries": [ "United States" ], "date": "2016-11-04", "year": 2016, "decimalYear": 2016.84153, "domains": [ "Language" ], "task": "Language modeling", "parameters": 51000000, "computeFlop": 21300000000000000, "log10Compute": 16.328, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1611.01462" }, { "name": "NASv3 (CIFAR-10)", "organization": "Google Brain", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2016-11-05", "year": 2016, "decimalYear": 2016.844262, "domains": [ "Vision" ], "task": "Image classification,Neural Architecture Search - NAS", "parameters": 37400000, "computeFlop": 2.2e+21, "log10Compute": 21.342, "trainingCostUsd2023": 21184.441090423978, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://arxiv.org/abs/1611.01578" }, { "name": "NAS with base 8 and shared embeddings", "organization": "Google Brain", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2016-11-05", "year": 2016, "decimalYear": 2016.844262, "domains": [ "Language" ], "task": "Language modeling,Neural Architecture Search - NAS", "parameters": 54000000, "computeFlop": 10500000000000000, "log10Compute": 16.021, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1611.01578" }, { "name": "BIDAF", "organization": "University of Washington,Allen Institute for AI", "orgCategory": "Academia,Research collective", "country": "United States", "countries": [ "United States" ], "date": "2016-11-05", "year": 2016, "decimalYear": 2016.844262, "domains": [ "Language" ], "task": "Question answering", "parameters": 2600000, "computeFlop": 3468614400000000000, "log10Compute": 18.54, "trainingCostUsd2023": 41.25014556779628, "powerDrawW": 4200.468649711526, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/1611.01603v6" }, { "name": "ResNeXt-101 (64×4d)", "organization": "University of California San Diego,Facebook", "orgCategory": "Academia,Industry", "country": "United States", "countries": [ "United States" ], "date": "2016-11-16", "year": 2016, "decimalYear": 2016.874317, "domains": [ "Vision" ], "task": "Image classification", "parameters": 83000000, "computeFlop": 12000000000000000000, "log10Compute": 19.079, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/1611.05431" }, { "name": "PolyNet", "organization": "Chinese University of Hong Kong (CUHK)", "orgCategory": "Academia", "country": "Hong Kong", "countries": [ "Hong Kong" ], "date": "2016-11-17", "year": 2016, "decimalYear": 2016.877049, "domains": [ "Vision" ], "task": "Image classification", "parameters": 92000000, "computeFlop": 64000000000000000000, "log10Compute": 19.806, "trainingCostUsd2023": 617.1106542, "powerDrawW": 16797.38519252759, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://arxiv.org/abs/1611.05725" }, { "name": "HR-ResNet101", "organization": "Carnegie Mellon University (CMU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2016-12-13", "year": 2016, "decimalYear": 2016.948087, "domains": [ "Vision" ], "task": "Face detection", "parameters": 44500000, "computeFlop": 7077000000000000000, "log10Compute": 18.85, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/Finding-Tiny-Faces-Hu-Ramanan/71f51e1b6691343ed031c1ed42efc96a7f1a0619" }, { "name": "EnhanceNet", "organization": "Max Planck Institute for Intelligent Systems", "orgCategory": "Academia", "country": "Germany", "countries": [ "Germany" ], "date": "2016-12-23", "year": 2016, "decimalYear": 2016.97541, "domains": [ "Vision" ], "task": "Image super-resolution", "parameters": 814464, "computeFlop": 130792319999999980, "log10Compute": 17.117, "trainingCostUsd2023": null, "powerDrawW": 282.4218009310853, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://www.semanticscholar.org/paper/EnhanceNet%3A-Single-Image-Super-Resolution-Through-Sajjadi-Scholkopf/fddc32f3880688238847077fd927ab3025db7a6a" }, { "name": "DeepStack", "organization": "University of Alberta,Charles University,Czech Technical University", "orgCategory": "Academia,Academia,Academia", "country": "Canada, Czechia", "countries": [ "Canada", "Czechia" ], "date": "2017-01-06", "year": 2017, "decimalYear": 2017.013699, "domains": [ "Games" ], "task": "Poker", "parameters": 2500000, "computeFlop": 14463360000000000000, "log10Compute": 19.16, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://arxiv.org/abs/1701.01724" }, { "name": "MoE-Multi", "organization": "Jagiellonian University,Google Brain", "orgCategory": "Academia,Industry", "country": "Poland, United States", "countries": [ "Poland", "United States" ], "date": "2017-01-23", "year": 2017, "decimalYear": 2017.060274, "domains": [ "Language" ], "task": "Language modeling,Translation", "parameters": 8700000000, "computeFlop": 93939056640000000000, "log10Compute": 19.973, "trainingCostUsd2023": 3874.12265, "powerDrawW": 32873.78910003616, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1701.06538" }, { "name": "Transformer", "organization": "Google Research,Google Brain", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2017-06-12", "year": 2017, "decimalYear": 2017.443836, "domains": [ "Language" ], "task": "Translation", "parameters": 213000000, "computeFlop": 7424524800000000000, "log10Compute": 18.871, "trainingCostUsd2023": 438.0356518424336, "powerDrawW": 4180.032869488383, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://proceedings.neurips.cc/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf" }, { "name": "DeepLoc", "organization": "Technical University of Denmark,University of Copenhagen", "orgCategory": "Academia,Academia", "country": "Denmark", "countries": [ "Denmark" ], "date": "2017-07-07", "year": 2017, "decimalYear": 2017.512329, "domains": [ "Biology" ], "task": "Protein localization prediction", "parameters": null, "computeFlop": 578102400000000000, "log10Compute": 17.762, "trainingCostUsd2023": null, "powerDrawW": 286.9303811522887, "frontier": false, "openWeights": false, "accessibility": "Hosted access (no API)", "confidence": "Likely", "link": "https://academic.oup.com/bioinformatics/article/33/21/3387/3931857" }, { "name": "JFT", "organization": "Google Research,Carnegie Mellon University (CMU)", "orgCategory": "Industry,Academia", "country": "United States", "countries": [ "United States" ], "date": "2017-07-10", "year": 2017, "decimalYear": 2017.520548, "domains": [ "Vision" ], "task": "Image classification,Object detection,Semantic segmentation,Pose estimation", "parameters": 44654504, "computeFlop": 843000000000000000000, "log10Compute": 20.926, "trainingCostUsd2023": 17910.759507843868, "powerDrawW": 31330.704406642573, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1707.02968" }, { "name": "ConvS2S (ensemble of 8 models)", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2017-07-25", "year": 2017, "decimalYear": 2017.561644, "domains": [ "Language" ], "task": "Translation", "parameters": null, "computeFlop": 56400000000000000000, "log10Compute": 19.751, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://arxiv.org/abs/1705.03122" }, { "name": "RetinaNet-R101", "organization": "Facebook AI Research", "orgCategory": "Industry", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2017-08-07", "year": 2017, "decimalYear": 2017.59726, "domains": [ "Vision" ], "task": "Object detection", "parameters": 53000000, "computeFlop": 2065392000000000000, "log10Compute": 18.315, "trainingCostUsd2023": null, "powerDrawW": 4174.823263154792, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://arxiv.org/abs/1708.02002" }, { "name": "AWD-LSTM - 3-layer LSTM (tied) + continuous cache pointer (WT2)", "organization": "Salesforce Research", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2017-08-07", "year": 2017, "decimalYear": 2017.59726, "domains": [ "Language" ], "task": "Language modeling", "parameters": 33000000, "computeFlop": 297000000000000000, "log10Compute": 17.473, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1708.02182" }, { "name": "OpenAI TI7 DOTA 1v1", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2017-08-11", "year": 2017, "decimalYear": 2017.608219, "domains": [ "Games" ], "task": "Dota 2", "parameters": null, "computeFlop": 604609522259200200000, "log10Compute": 20.781, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": null, "accessibility": null, "confidence": "Confident", "link": "https://openai.com/research/dota-2" }, { "name": "EI-REHN-1000D", "organization": "Korea Advanced Institute of Science and Technology (KAIST)", "orgCategory": "Academia", "country": "South Korea", "countries": [ "South Korea" ], "date": "2017-08-14", "year": 2017, "decimalYear": 2017.616438, "domains": [ "Language" ], "task": "Language modeling", "parameters": 19000000, "computeFlop": 10600000000000000, "log10Compute": 16.025, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1708.04116" }, { "name": "Libratus", "organization": "Carnegie Mellon University (CMU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2017-08-19", "year": 2017, "decimalYear": 2017.630137, "domains": [ "Games" ], "task": "Poker", "parameters": null, "computeFlop": 551000000000000000000, "log10Compute": 20.741, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://www.ijcai.org/proceedings/2017/0772.pdf" }, { "name": "GL-LWGC-AWD-MoS-LSTM + dynamic evaluation (WT2)", "organization": "Ben-Gurion University of the Negev", "orgCategory": "Academia", "country": "Israel", "countries": [ "Israel" ], "date": "2017-08-29", "year": 2017, "decimalYear": 2017.657534, "domains": [ "Language" ], "task": "Language modeling", "parameters": 38000000, "computeFlop": 456000000000000000, "log10Compute": 17.659, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1708.08863" }, { "name": "PyramidNet", "organization": "Korea Advanced Institute of Science and Technology (KAIST)", "orgCategory": "Academia", "country": "South Korea", "countries": [ "South Korea" ], "date": "2017-09-06", "year": 2017, "decimalYear": 2017.679452, "domains": [ "Vision" ], "task": "Image classification", "parameters": 26000000, "computeFlop": 2340000000000000, "log10Compute": 15.369, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/1610.02915v4" }, { "name": "ISS", "organization": "Duke University,Microsoft", "orgCategory": "Academia,Industry", "country": "United States", "countries": [ "United States" ], "date": "2017-09-15", "year": 2017, "decimalYear": 2017.70411, "domains": [ "Language" ], "task": "Language modeling", "parameters": 11100000.000000002, "computeFlop": 3400000000000000, "log10Compute": 15.531, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1709.05027" }, { "name": "AWD-LSTM+WT+Cache+IOG (WT2)", "organization": "NTT Communication Science Laboratories", "orgCategory": "Industry", "country": "Japan", "countries": [ "Japan" ], "date": "2017-09-26", "year": 2017, "decimalYear": 2017.734247, "domains": [ "Language" ], "task": "Language modeling", "parameters": 53000000, "computeFlop": 3180000000000000, "log10Compute": 15.502, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1709.08907" }, { "name": "AlphaGo Zero", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2017-10-18", "year": 2017, "decimalYear": 2017.794521, "domains": [ "Games" ], "task": "Go", "parameters": 46400244, "computeFlop": 649439910290227000000, "log10Compute": 20.813, "trainingCostUsd2023": 1213.8707315707268, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://www.nature.com/articles/nature24270" }, { "name": "AlphaGo Master", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2017-10-19", "year": 2017, "decimalYear": 2017.79726, "domains": [ "Games" ], "task": "Go", "parameters": null, "computeFlop": 341000000010000000000, "log10Compute": 20.533, "trainingCostUsd2023": 471445.3247973037, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://www.nature.com/articles/nature24270" }, { "name": "Fraternal dropout + AWD-LSTM 3-layer (WT2)", "organization": "Jagiellonian University,Mila - Quebec AI (originally Montreal Institute for Learning Algorithms),University of Montreal / Université de Montréal", "orgCategory": "Academia,Academia,Academia", "country": "Poland, Canada", "countries": [ "Poland", "Canada" ], "date": "2017-10-31", "year": 2017, "decimalYear": 2017.830137, "domains": [ "Language" ], "task": "Language modeling", "parameters": 34000000, "computeFlop": 306000000000000000, "log10Compute": 17.486, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1711.00066" }, { "name": "AWD-LSTM-MoS + dynamic evaluation (WT2, 2017)", "organization": "Carnegie Mellon University (CMU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2017-11-10", "year": 2017, "decimalYear": 2017.857534, "domains": [ "Language" ], "task": "Language modeling", "parameters": 35000000, "computeFlop": 3360000000000000000, "log10Compute": 18.526, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1711.03953" }, { "name": "AlphaZero", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2017-12-05", "year": 2017, "decimalYear": 2017.926027, "domains": [ "Games" ], "task": "Chess,Shogi,Go", "parameters": null, "computeFlop": 106000000010000000000, "log10Compute": 20.025, "trainingCostUsd2023": 229918.6146969874, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1712.01815" }, { "name": "QRNN", "organization": "Salesforce Research", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2018-02-01", "year": 2018, "decimalYear": 2018.084932, "domains": [ "Language" ], "task": "Language modeling", "parameters": 135000000, "computeFlop": 688664720000000000, "log10Compute": 17.838, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://mlsys.org/Conferences/doc/2018/50.pdf" }, { "name": "ELMo", "organization": "University of Washington,Allen Institute for AI", "orgCategory": "Academia,Research collective", "country": "United States", "countries": [ "United States" ], "date": "2018-02-01", "year": 2018, "decimalYear": 2018.084932, "domains": [ "Language" ], "task": "Question answering,Sentiment classification,Language modeling", "parameters": 94000000, "computeFlop": 3300100000000010, "log10Compute": 15.519, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Speculative", "link": "https://arxiv.org/abs/1802.05365" }, { "name": "IMPALA", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2018-02-05", "year": 2018, "decimalYear": 2018.09589, "domains": [ "Games" ], "task": "Atari", "parameters": 1600000, "computeFlop": 168000000000000000000, "log10Compute": 20.225, "trainingCostUsd2023": 53.42846804494412, "powerDrawW": 285.5725852593349, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1802.01561" }, { "name": "4 layer QRNN (h=2500)", "organization": "Salesforce Research", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2018-03-22", "year": 2018, "decimalYear": 2018.219178, "domains": [ "Language" ], "task": "Language modeling", "parameters": 151000000, "computeFlop": 591588150000000000, "log10Compute": 17.772, "trainingCostUsd2023": null, "powerDrawW": 268.1693571105067, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1803.08240" }, { "name": "YOLOv3", "organization": "University of Washington", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2018-04-08", "year": 2018, "decimalYear": 2018.265753, "domains": [ "Vision" ], "task": "Object detection", "parameters": 56933216, "computeFlop": 13416380824000000000, "log10Compute": 19.128, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1804.02767" }, { "name": "ResNeXt-101 32x48d", "organization": "Facebook", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2018-05-02", "year": 2018, "decimalYear": 2018.331507, "domains": [ "Vision" ], "task": "Image classification", "parameters": 829000000, "computeFlop": 8.74395e+21, "log10Compute": 21.942, "trainingCostUsd2023": 134077.42688552555, "powerDrawW": 209159.05867416784, "frontier": true, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/1805.00932" }, { "name": "Dropout-LSTM+Noise(Bernoulli) (WT2)", "organization": "Columbia University,New York University (NYU),Princeton University", "orgCategory": "Academia,Academia,Academia", "country": "United States", "countries": [ "United States" ], "date": "2018-05-03", "year": 2018, "decimalYear": 2018.334247, "domains": [ "Language" ], "task": "Language modeling", "parameters": 51000000, "computeFlop": 127000000000000000, "log10Compute": 17.104, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1805.01500" }, { "name": "aLSTM(depth-2)+RecurrentPolicy (WT2)", "organization": "University of Manchester,Alan Turing Institute", "orgCategory": "Academia,Government", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2018-05-22", "year": 2018, "decimalYear": 2018.386301, "domains": [ "Language" ], "task": "Language modeling", "parameters": 32000000, "computeFlop": 72960000000000000, "log10Compute": 16.863, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1805.08574" }, { "name": "GPT-1", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2018-06-01", "year": 2018, "decimalYear": 2018.413699, "domains": [ "Language" ], "task": "Question answering,Text classification,Language modeling", "parameters": 117000000, "computeFlop": 17578125000000000000, "log10Compute": 19.245, "trainingCostUsd2023": null, "powerDrawW": 663.5535559333689, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://openai.com/blog/language-unsupervised/" }, { "name": "FTW (For The Win)", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2018-07-03", "year": 2018, "decimalYear": 2018.50137, "domains": [ "Games" ], "task": "Capture the flag", "parameters": 126001330, "computeFlop": 34900000000000000000, "log10Compute": 19.543, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://arxiv.org/abs/1807.01281" }, { "name": "Big-Little Net (speech)", "organization": "IBM", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2018-07-10", "year": 2018, "decimalYear": 2018.520548, "domains": [ "Speech" ], "task": "Speech recognition (ASR)", "parameters": 3320000, "computeFlop": 429004800000000000, "log10Compute": 17.632, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Speculative", "link": "https://arxiv.org/abs/1807.03848" }, { "name": "Big-Little Net", "organization": "IBM", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2018-07-10", "year": 2018, "decimalYear": 2018.520548, "domains": [ "Vision" ], "task": "Image classification,Object recognition", "parameters": 77360000, "computeFlop": 246048000000000000, "log10Compute": 17.391, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/1807.03848" }, { "name": "Big Transformer for Back-Translation", "organization": "Facebook AI Research,Google Brain", "orgCategory": "Industry,Industry", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2018-08-28", "year": 2018, "decimalYear": 2018.654795, "domains": [ "Language" ], "task": "Translation", "parameters": null, "computeFlop": 478080000000000000000, "log10Compute": 20.68, "trainingCostUsd2023": 2442.1618775733145, "powerDrawW": 66225.44602672113, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/1808.09381" }, { "name": "(ensemble): AWD-LSTM-DOC (fin) × 5 (WT2)", "organization": "NTT Communication Science Laboratories,Tohoku University", "orgCategory": "Industry,Academia", "country": "Japan", "countries": [ "Japan" ], "date": "2018-08-30", "year": 2018, "decimalYear": 2018.660274, "domains": [ "Language" ], "task": "Language modeling", "parameters": 185000000, "computeFlop": 666000000000000000, "log10Compute": 17.823, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/1808.10143" }, { "name": "Transformer + Simple Recurrent Unit", "organization": "ASAPP,Cornell University,Google,Princeton University", "orgCategory": "Industry,Academia,Industry,Academia", "country": "United States", "countries": [ "United States" ], "date": "2018-09-17", "year": 2018, "decimalYear": 2018.709589, "domains": [ "Language" ], "task": "Translation", "parameters": 90000000, "computeFlop": 11000000000000000000, "log10Compute": 19.041, "trainingCostUsd2023": 45.373219244858774, "powerDrawW": 4964.696746005496, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1709.02755v5" }, { "name": "LSTM+NeuralCache", "organization": "KU Leuven,ESAT - PSI,Apple", "orgCategory": "Academia,Academia,Industry", "country": "Belgium, United States", "countries": [ "Belgium", "United States" ], "date": "2018-09-24", "year": 2018, "decimalYear": 2018.728767, "domains": [ "Language" ], "task": "Language modeling", "parameters": 2100000, "computeFlop": 982800000000000, "log10Compute": 14.992, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1809.08826" }, { "name": "Transformer (Adaptive Input Embeddings) WT103", "organization": "Facebook AI Research", "orgCategory": "Industry", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2018-09-28", "year": 2018, "decimalYear": 2018.739726, "domains": [ "Language" ], "task": "Language modeling", "parameters": 247000000, "computeFlop": 44700000000000000000, "log10Compute": 19.65, "trainingCostUsd2023": 2880.917278699733, "powerDrawW": 4963.480727525225, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/1809.10853" }, { "name": "BERT-Large", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2018-10-11", "year": 2018, "decimalYear": 2018.775342, "domains": [ "Language" ], "task": "Question answering,Text autocompletion", "parameters": 340000000, "computeFlop": 285000000000000000000, "log10Compute": 20.455, "trainingCostUsd2023": 1751.4770087736404, "powerDrawW": 37049.928524121315, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/1810.04805" }, { "name": "TrellisNet", "organization": "Carnegie Mellon University (CMU),Bosch Center for Artificial Intelligence,Intel Labs", "orgCategory": "Academia,Industry,Industry", "country": "United States, Germany", "countries": [ "United States", "Germany" ], "date": "2018-10-15", "year": 2018, "decimalYear": 2018.786301, "domains": [ "Language" ], "task": "Language modeling", "parameters": 180000000, "computeFlop": 2780000000000000000, "log10Compute": 18.444, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1810.06682" }, { "name": "Mesh-TensorFlow Transformer 4.9B (language)", "organization": "Google Brain", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2018-11-05", "year": 2018, "decimalYear": 2018.843836, "domains": [ "Language" ], "task": "Language modeling/generation,Translation", "parameters": 4900000000, "computeFlop": 161740800000000000000, "log10Compute": 20.209, "trainingCostUsd2023": 935.3300509163912, "powerDrawW": 148117.2291982981, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1811.02084" }, { "name": "Mesh-TensorFlow Transformer 2.9B (translation)", "organization": "Google Brain", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2018-11-05", "year": 2018, "decimalYear": 2018.843836, "domains": [ "Language" ], "task": "Language modeling/generation,Translation", "parameters": 2900000000, "computeFlop": 68428800000000000000, "log10Compute": 19.835, "trainingCostUsd2023": 395.71656000308855, "powerDrawW": 37029.307299574524, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1811.02084" }, { "name": "Fine-tuned-AWD-LSTM-DOC (fin)", "organization": "Samsung R&D Institute Russia", "orgCategory": "Industry", "country": "Russia", "countries": [ "Russia" ], "date": "2018-11-12", "year": 2018, "decimalYear": 2018.863014, "domains": [ "Language" ], "task": "Language modeling", "parameters": 46000000, "computeFlop": 51880000000000000, "log10Compute": 16.715, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1811.04623" }, { "name": "Multi-cell LSTM", "organization": "University of Hyderabad", "orgCategory": "Academia", "country": "India", "countries": [ "India" ], "date": "2018-11-15", "year": 2018, "decimalYear": 2018.871233, "domains": [ "Language" ], "task": "Language modeling", "parameters": 7200000, "computeFlop": 2006640000000000, "log10Compute": 15.302, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1811.06477" }, { "name": "StyleGAN", "organization": "NVIDIA", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2018-12-12", "year": 2018, "decimalYear": 2018.945205, "domains": [ "Image generation" ], "task": "Image generation", "parameters": 26200000, "computeFlop": 39300000000000000, "log10Compute": 16.594, "trainingCostUsd2023": null, "powerDrawW": 4955.197627044293, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/1812.04948" }, { "name": "Transformer-XL (257M)", "organization": "Carnegie Mellon University (CMU),Google Brain", "orgCategory": "Academia,Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-01-09", "year": 2019, "decimalYear": 2019.021918, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 256999999.99999997, "computeFlop": 378327710000000000000, "log10Compute": 20.578, "trainingCostUsd2023": null, "powerDrawW": 29712.65288608572, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Speculative", "link": "https://arxiv.org/abs/1901.02860" }, { "name": "Hanabi 4 player", "organization": "DeepMind,University of Oxford,Carnegie Mellon University (CMU),Google Brain", "orgCategory": "Industry,Academia,Academia,Industry", "country": "United Kingdom, United States", "countries": [ "United Kingdom", "United States" ], "date": "2019-02-01", "year": 2019, "decimalYear": 2019.084932, "domains": [ "Games" ], "task": "Hanabi", "parameters": 764000, "computeFlop": 4300000000000000000, "log10Compute": 18.633, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1902.00506" }, { "name": "GPT-2 (1.5B)", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-02-14", "year": 2019, "decimalYear": 2019.120548, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 1500000000, "computeFlop": 1.920000000001e+21, "log10Compute": 21.283, "trainingCostUsd2023": 4348.443645, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Speculative", "link": "https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf" }, { "name": "KataGo", "organization": "Jane Street", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-02-27", "year": 2019, "decimalYear": 2019.156164, "domains": [ "Games" ], "task": "Go", "parameters": 2500000, "computeFlop": 23200000000000000000, "log10Compute": 19.365, "trainingCostUsd2023": 104.91425851608678, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Speculative", "link": "https://arxiv.org/abs/1902.10565" }, { "name": "SciBERT", "organization": "Allen Institute for AI", "orgCategory": "Research collective", "country": "United States", "countries": [ "United States" ], "date": "2019-03-26", "year": 2019, "decimalYear": 2019.230137, "domains": [ "Language" ], "task": "Relation extraction,Sentiment classification,Text classification,Named entity recognition (NER)", "parameters": 110000000, "computeFlop": 89268480000000000000, "log10Compute": 19.951, "trainingCostUsd2023": 247.26289010271603, "powerDrawW": 3707.8009472465747, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/1903.10676" }, { "name": "Cross-lingual alignment", "organization": "Tel Aviv University,Massachusetts Institute of Technology (MIT)", "orgCategory": "Academia,Academia", "country": "Israel, United States", "countries": [ "Israel", "United States" ], "date": "2019-04-04", "year": 2019, "decimalYear": 2019.254795, "domains": [ "Language" ], "task": "Translation", "parameters": null, "computeFlop": 2560000000000000000, "log10Compute": 18.408, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Speculative", "link": "https://arxiv.org/abs/1902.09492" }, { "name": "WeNet (Penn Treebank)", "organization": "Amazon", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-04-08", "year": 2019, "decimalYear": 2019.265753, "domains": [ "Language" ], "task": "Neural Architecture Search - NAS,Language modeling", "parameters": 23000000, "computeFlop": 730000001000000000, "log10Compute": 17.863, "trainingCostUsd2023": null, "powerDrawW": 339.4439274233134, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1904.03819" }, { "name": "BERT-Large-CAS (PTB+WT2+WT103)", "organization": "Amazon", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-04-20", "year": 2019, "decimalYear": 2019.29863, "domains": [ "Language" ], "task": "Neural Architecture Search - NAS,Language modeling/generation", "parameters": 395000000, "computeFlop": 154050000000000000000, "log10Compute": 20.188, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1904.09408" }, { "name": "MuseNet", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-04-25", "year": 2019, "decimalYear": 2019.312329, "domains": [ "Audio" ], "task": "Audio generation", "parameters": 2038431744, "computeFlop": 220830105600000000000, "log10Compute": 20.344, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://openai.com/index/musenet/" }, { "name": "AWD-LSTM-DRILL + dynamic evaluation† (WT2)", "organization": "IDIAP", "orgCategory": "Academia", "country": "Switzerland", "countries": [ "Switzerland" ], "date": "2019-05-14", "year": 2019, "decimalYear": 2019.364384, "domains": [ "Language" ], "task": "Language modeling", "parameters": 34000000, "computeFlop": 408000000000000000, "log10Compute": 17.611, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/1905.05513" }, { "name": "DLRM-2020", "organization": "Facebook AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-05-31", "year": 2019, "decimalYear": 2019.410959, "domains": [ "Recommendation" ], "task": "Recommender system", "parameters": 100000000000, "computeFlop": 4000000000000000000, "log10Compute": 18.602, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1906.00091" }, { "name": "XLNet", "organization": "Carnegie Mellon University (CMU),Google Brain", "orgCategory": "Academia,Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-06-01", "year": 2019, "decimalYear": 2019.413699, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Sentiment classification", "parameters": 340000000, "computeFlop": 6.19e+21, "log10Compute": 21.792, "trainingCostUsd2023": 13644.959135438661, "powerDrawW": null, "frontier": true, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/1906.08237" }, { "name": "Transformer-XL Large + Phrase Induction", "organization": "Massachusetts Institute of Technology (MIT),University of Illinois Urbana-Champaign (UIUC)", "orgCategory": "Academia,Academia", "country": "United States", "countries": [ "United States" ], "date": "2019-06-04", "year": 2019, "decimalYear": 2019.421918, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 256999999.99999997, "computeFlop": 378486510000000000000, "log10Compute": 20.578, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://arxiv.org/abs/1906.01702" }, { "name": "AWD-LSTM + MoS + Partial Shuffled", "organization": "University of Texas at Austin", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2019-06-10", "year": 2019, "decimalYear": 2019.438356, "domains": [ "Language" ], "task": "Language modeling", "parameters": 35000000, "computeFlop": 315000000000000000, "log10Compute": 17.498, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/1906.03805" }, { "name": "RoBERTa Large", "organization": "Facebook,University of Washington", "orgCategory": "Industry,Academia", "country": "United States", "countries": [ "United States" ], "date": "2019-07-01", "year": 2019, "decimalYear": 2019.49589, "domains": [ "Language" ], "task": "Question answering,Language modeling/generation", "parameters": 355000000, "computeFlop": 8.5067e+21, "log10Compute": 21.93, "trainingCostUsd2023": 85350.2350127364, "powerDrawW": 526193.8152119832, "frontier": true, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/1907.11692" }, { "name": "Pluribus", "organization": "Facebook AI Research", "orgCategory": "Industry", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2019-07-11", "year": 2019, "decimalYear": 2019.523288, "domains": [ "Games" ], "task": "Poker", "parameters": null, "computeFlop": 66000000000000000, "log10Compute": 16.82, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://www.science.org/cms/asset/910714a7-ee2a-486e-9970-42fb893b08d9/pap.pdf" }, { "name": "trRosetta", "organization": "Nankai University,University of Washington,Tianjin University,Harvard University", "orgCategory": "Academia,Academia,Academia,Academia", "country": "China, United States", "countries": [ "China", "United States" ], "date": "2019-08-22", "year": 2019, "decimalYear": 2019.638356, "domains": [ "Biology" ], "task": "Protein folding prediction", "parameters": null, "computeFlop": 38047968000000000000, "log10Compute": 19.58, "trainingCostUsd2023": null, "powerDrawW": 315.8562689911153, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://www.pnas.org/doi/10.1073/pnas.1914677117" }, { "name": "UDSMProt", "organization": "Fraunhofer Heinrich Hertz Institute", "orgCategory": "Research collective", "country": "Germany", "countries": [ "Germany" ], "date": "2019-09-04", "year": 2019, "decimalYear": 2019.673973, "domains": [ "Biology" ], "task": "Proteins,Protein or nucleotide language model (pLM/nLM),Enzyme function prediction", "parameters": 28303800, "computeFlop": 637000000000000000, "log10Compute": 17.804, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://www.biorxiv.org/content/10.1101/704874v2.full.pdf" }, { "name": "Megatron-LM (8.3B)", "organization": "NVIDIA", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-09-17", "year": 2019, "decimalYear": 2019.709589, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 8300000000, "computeFlop": 9.1e+21, "log10Compute": 21.959, "trainingCostUsd2023": 109287.55758939347, "powerDrawW": 262640.30207328824, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1909.08053" }, { "name": "Megatron-LM (1.2B)", "organization": "NVIDIA", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-09-17", "year": 2019, "decimalYear": 2019.709589, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 1200000000, "computeFlop": 1.1304e+21, "log10Compute": 21.053, "trainingCostUsd2023": null, "powerDrawW": 338.2215428484996, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1909.08053" }, { "name": "Megatron-BERT", "organization": "NVIDIA", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-09-17", "year": 2019, "decimalYear": 2019.709589, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 3900000000, "computeFlop": 2.2e+22, "log10Compute": 22.342, "trainingCostUsd2023": 171818.64196707975, "powerDrawW": 262640.30207328824, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1909.08053" }, { "name": "DistilBERT", "organization": "Hugging Face", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-10-02", "year": 2019, "decimalYear": 2019.750685, "domains": [ "Language" ], "task": "Text autocompletion", "parameters": 66000000, "computeFlop": 12441600000000000000, "log10Compute": 19.095, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/1910.01108" }, { "name": "AlphaX-1", "organization": "Facebook AI Research,Brown University", "orgCategory": "Industry,Academia", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2019-10-02", "year": 2019, "decimalYear": 2019.750685, "domains": [ "Vision" ], "task": "Neural architecture search for computer vision,Image classification,Object detection,Image captioning", "parameters": 5400000, "computeFlop": 889344000000000000, "log10Compute": 17.949, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1903.11059" }, { "name": "T5-11B", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-10-23", "year": 2019, "decimalYear": 2019.808219, "domains": [ "Language" ], "task": "Text autocompletion,Language modeling/generation", "parameters": 11000000000, "computeFlop": 3.3e+22, "log10Compute": 22.519, "trainingCostUsd2023": 78464.68282814459, "powerDrawW": 472373.69114573154, "frontier": true, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/1910.10683" }, { "name": "T5-3B", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-10-23", "year": 2019, "decimalYear": 2019.808219, "domains": [ "Language" ], "task": "Text autocompletion", "parameters": 2800000000, "computeFlop": 9.0000000001e+21, "log10Compute": 21.954, "trainingCostUsd2023": 17420.709393371122, "powerDrawW": null, "frontier": true, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/1910.10683" }, { "name": "AlphaStar", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2019-10-30", "year": 2019, "decimalYear": 2019.827397, "domains": [ "Games" ], "task": "StarCraft", "parameters": 139000000, "computeFlop": 1.0773400001e+23, "log10Compute": 23.032, "trainingCostUsd2023": 130654.07330431018, "powerDrawW": 354225.04547181545, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://www.deepmind.com/blog/alphastar-grandmaster-level-in-starcraft-ii-using-multi-agent-reinforcement-learning" }, { "name": "Base LM + kNN LM + Continuous Cache", "organization": "Stanford University,Facebook AI Research", "orgCategory": "Academia,Industry", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2019-11-01", "year": 2019, "decimalYear": 2019.832877, "domains": [ "Language" ], "task": "Language modeling", "parameters": 247000000.00000003, "computeFlop": 30529200000000000000, "log10Compute": 19.485, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/1911.00172" }, { "name": "XLM-RoBERTa", "organization": "Facebook AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-11-05", "year": 2019, "decimalYear": 2019.843836, "domains": [ "Language" ], "task": "Named entity recognition (NER),Question answering,Text classification", "parameters": 550000000, "computeFlop": 2.076e+22, "log10Compute": 22.317, "trainingCostUsd2023": 77861.31819548723, "powerDrawW": 256204.94677329363, "frontier": true, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/1911.02116" }, { "name": "Sandwich Transformer", "organization": "Allen Institute for AI,Facebook AI Research", "orgCategory": "Research collective,Industry", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2019-11-10", "year": 2019, "decimalYear": 2019.857534, "domains": [ "Language" ], "task": "Language modeling", "parameters": 209000000, "computeFlop": 23504093000000000000, "log10Compute": 19.371, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://arxiv.org/abs/1911.03864" }, { "name": "CamemBERT", "organization": "Facebook,INRIA,Sorbonne University", "orgCategory": "Industry,Academia,Academia", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2019-11-10", "year": 2019, "decimalYear": 2019.857534, "domains": [ "Language" ], "task": "Language modeling/generation,Part-of-speech tagging,Named entity recognition (NER)", "parameters": 335000000, "computeFlop": 830000000000000000000, "log10Compute": 20.919, "trainingCostUsd2023": 2319.5419478533995, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/1911.03894" }, { "name": "Noisy Student (L2)", "organization": "Carnegie Mellon University (CMU),Google", "orgCategory": "Academia,Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-11-11", "year": 2019, "decimalYear": 2019.860274, "domains": [ "Vision" ], "task": "Image classification", "parameters": 480000000, "computeFlop": 2.612e+22, "log10Compute": 22.417, "trainingCostUsd2023": 45609.73119622589, "powerDrawW": 944347.7271739373, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1911.04252v4" }, { "name": "MuZero", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2019-11-19", "year": 2019, "decimalYear": 2019.882192, "domains": [ "Games" ], "task": "Atari", "parameters": 36864000, "computeFlop": 48000000000000000000, "log10Compute": 19.681, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://arxiv.org/abs/1911.08265v2" }, { "name": "Transformer-XL DeFINE (141M)", "organization": "University of Washington,Allen Institute for AI", "orgCategory": "Academia,Research collective", "country": "United States", "countries": [ "United States" ], "date": "2019-11-27", "year": 2019, "decimalYear": 2019.90411, "domains": [ "Language" ], "task": "Language modeling", "parameters": 141000000, "computeFlop": 1742760000000000000, "log10Compute": 18.241, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://arxiv.org/abs/1911.12385" }, { "name": "MMLSTM (WT-2)", "organization": "Beijing University of Posts and Telecommunications,University of West London", "orgCategory": "Academia,Academia", "country": "China, United Kingdom", "countries": [ "China", "United Kingdom" ], "date": "2019-12-05", "year": 2019, "decimalYear": 2019.926027, "domains": [ "Language" ], "task": "Language modeling", "parameters": 32300000, "computeFlop": 193800000000000000, "log10Compute": 17.287, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "http://repository.uwl.ac.uk/id/eprint/6490/1/Loo_etal_IEEE_TNNLS_2019_Major-minor_long_short-term_memory_for_word-level_language_model.pdf" }, { "name": "MMLSTM (PTB)", "organization": "Beijing University of Posts and Telecommunications,University of West London", "orgCategory": "Academia,Academia", "country": "China, United Kingdom", "countries": [ "China", "United Kingdom" ], "date": "2019-12-05", "year": 2019, "decimalYear": 2019.926027, "domains": [ "Language" ], "task": "Language modeling", "parameters": 21300000, "computeFlop": 58298782000000000, "log10Compute": 16.766, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "http://repository.uwl.ac.uk/id/eprint/6490/1/Loo_etal_IEEE_TNNLS_2019_Major-minor_long_short-term_memory_for_word-level_language_model.pdf" }, { "name": "OpenAI Five Rerun", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-12-13", "year": 2019, "decimalYear": 2019.947945, "domains": [ "Games" ], "task": "Dota 2", "parameters": 159000000, "computeFlop": 1.3e+22, "log10Compute": 22.114, "trainingCostUsd2023": 321105.9055, "powerDrawW": 262131.94609302413, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://cdn.openai.com/dota-2.pdf" }, { "name": "OpenAI Five", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2019-12-13", "year": 2019, "decimalYear": 2019.947945, "domains": [ "Games" ], "task": "Dota 2", "parameters": 159000000, "computeFlop": 6.7e+22, "log10Compute": 22.826, "trainingCostUsd2023": 4328311.470477086, "powerDrawW": 786395.8382790724, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1912.06680" }, { "name": "DD-PPO", "organization": "Georgia Institute of Technology,Facebook AI Research,Oregon State University,Simon Fraser University", "orgCategory": "Academia,Industry,Academia,Academia", "country": "United States, France, Canada", "countries": [ "United States", "France", "Canada" ], "date": "2019-12-19", "year": 2019, "decimalYear": 2019.964384, "domains": [ "Robotics" ], "task": "Object detection", "parameters": null, "computeFlop": 780000000000000000000, "log10Compute": 20.892, "trainingCostUsd2023": 1926.8992900057376, "powerDrawW": 39314.53850266432, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://openreview.net/forum?id=H1gX8C4YPr\nhttps://arxiv.org/abs/1911.00357" }, { "name": "AlphaFold", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2020-01-15", "year": 2020, "decimalYear": 2020.038251, "domains": [ "Biology" ], "task": "Protein folding prediction,Proteins", "parameters": 16340840, "computeFlop": 100000000000000000000, "log10Compute": 20, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://www.nature.com/articles/s41586-019-1923-7" }, { "name": "ContextNet + Noisy Student", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-01-19", "year": 2020, "decimalYear": 2020.04918, "domains": [ "Speech" ], "task": "Speech recognition (ASR)", "parameters": null, "computeFlop": 8.16e+21, "log10Compute": 21.912, "trainingCostUsd2023": 14226.054462994534, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2005.09629v2" }, { "name": "Meena", "organization": "Google Brain", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-01-28", "year": 2020, "decimalYear": 2020.07377, "domains": [ "Language" ], "task": "Text autocompletion,Chat", "parameters": 2600000000, "computeFlop": 1.12e+23, "log10Compute": 23.049, "trainingCostUsd2023": 214809.91213508433, "powerDrawW": 942708.8086440804, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2001.09977" }, { "name": "TaLK Convolution", "organization": "Carleton University", "orgCategory": "Academia", "country": "Canada", "countries": [ "Canada" ], "date": "2020-02-08", "year": 2020, "decimalYear": 2020.103825, "domains": [ "Language" ], "task": "Language modeling,Translation,Text summarization", "parameters": 240000000, "computeFlop": 26990346000000000000, "log10Compute": 19.431, "trainingCostUsd2023": null, "powerDrawW": 4090.615920439178, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2002.03184" }, { "name": "ALBERT-xxlarge", "organization": "Toyota Technological Institute at Chicago,Google", "orgCategory": "Academia,Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-02-09", "year": 2020, "decimalYear": 2020.106557, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering", "parameters": 235000000, "computeFlop": 2.39e+21, "log10Compute": 21.378, "trainingCostUsd2023": 4439.921298510215, "powerDrawW": 471228.45995621517, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Speculative", "link": "https://arxiv.org/abs/1909.11942" }, { "name": "Turing-NLG", "organization": "Microsoft", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-02-13", "year": 2020, "decimalYear": 2020.117486, "domains": [ "Language" ], "task": "Text autocompletion,Language generation,Text summarization", "parameters": 17000000000, "computeFlop": 1.57e+22, "log10Compute": 22.196, "trainingCostUsd2023": 51659.713290894986, "powerDrawW": 130885.13499433055, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://www.microsoft.com/en-us/research/blog/turing-nlg-a-17-billion-parameter-language-model-by-microsoft/" }, { "name": "FFN SwiGLU", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-02-14", "year": 2020, "decimalYear": 2020.120219, "domains": [ "Language" ], "task": "Language modeling,Question answering", "parameters": 220000000, "computeFlop": 33687317000000000000, "log10Compute": 19.527, "trainingCostUsd2023": null, "powerDrawW": 9161.75542079835, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2002.05202" }, { "name": "Feedback Transformer", "organization": "LORIA,University of Lorraine,Facebook AI Research", "orgCategory": "Academia,Academia,Industry", "country": "France, United States", "countries": [ "France", "United States" ], "date": "2020-02-21", "year": 2020, "decimalYear": 2020.139344, "domains": [ "Language" ], "task": "Language modeling", "parameters": 126000000, "computeFlop": 7690547000000000000, "log10Compute": 18.886, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://arxiv.org/abs/2002.09402" }, { "name": "TransformerXL + spectrum control", "organization": "University of California Los Angeles (UCLA),JD.com", "orgCategory": "Academia,Industry", "country": "United States, China", "countries": [ "United States", "China" ], "date": "2020-03-11", "year": 2020, "decimalYear": 2020.191257, "domains": [ "Language" ], "task": "Language modeling", "parameters": 151000000, "computeFlop": 26289761000000000000, "log10Compute": 19.42, "trainingCostUsd2023": null, "powerDrawW": 2452.6211427753756, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://openreview.net/forum?id=ByxY8CNtvr" }, { "name": "Tensor-Transformer(1core)+PN (WT103)", "organization": "University of California (UC) Berkeley", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2020-03-17", "year": 2020, "decimalYear": 2020.20765, "domains": [ "Language" ], "task": "Language modeling", "parameters": 85300000, "computeFlop": 1580000000000000000, "log10Compute": 18.199, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2003.07845" }, { "name": "ELECTRA", "organization": "Stanford University,Google,Google Brain", "orgCategory": "Academia,Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-03-23", "year": 2020, "decimalYear": 2020.224044, "domains": [ "Language" ], "task": "Text autocompletion", "parameters": 335000000, "computeFlop": 3.1e+21, "log10Compute": 21.491, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2003.10555v1" }, { "name": "MetNet", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-03-24", "year": 2020, "decimalYear": 2020.226776, "domains": [ "Earth science" ], "task": "Weather forecasting", "parameters": 225000000, "computeFlop": 9510912000000000000, "log10Compute": 18.978, "trainingCostUsd2023": null, "powerDrawW": 235383.4758953672, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://arxiv.org/abs/2003.12140" }, { "name": "Once for All", "organization": "MIT-IBM Watson AI Lab,Massachusetts Institute of Technology (MIT),IBM", "orgCategory": "Academia,Industry,Academia,Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-04-29", "year": 2020, "decimalYear": 2020.325137, "domains": [ "Vision" ], "task": "Image classification", "parameters": 7700000, "computeFlop": 623700000000000000000, "log10Compute": 20.795, "trainingCostUsd2023": 1753.9255676777682, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/1908.09791" }, { "name": "UnifiedQA", "organization": "Allen Institute for AI,University of Washington", "orgCategory": "Research collective,Academia", "country": "United States", "countries": [ "United States" ], "date": "2020-05-02", "year": 2020, "decimalYear": 2020.333333, "domains": [ "Language" ], "task": "Question answering", "parameters": 11000000000, "computeFlop": 16500000000000000000, "log10Compute": 19.217, "trainingCostUsd2023": 59.122086907352475, "powerDrawW": 7349.34790019072, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2005.00700v3" }, { "name": "DETR", "organization": "Facebook", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-05-26", "year": 2020, "decimalYear": 2020.398907, "domains": [ "Vision" ], "task": "Object detection", "parameters": 60000000, "computeFlop": 400000000000000000000, "log10Compute": 20.602, "trainingCostUsd2023": 959.9097627206426, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2005.12872" }, { "name": "GPT-3 175B (davinci)", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-05-28", "year": 2020, "decimalYear": 2020.404372, "domains": [ "Language" ], "task": "Text autocompletion,Language modeling/generation", "parameters": 174600000000, "computeFlop": 3.14e+23, "log10Compute": 23.497, "trainingCostUsd2023": 2116866.2428390156, "powerDrawW": 5100759.605963441, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Confident", "link": "https://arxiv.org/abs/2005.14165" }, { "name": "GShard (dense)", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-06-30", "year": 2020, "decimalYear": 2020.494536, "domains": [ "Language" ], "task": "Translation", "parameters": 2300000000, "computeFlop": 4.765e+22, "log10Compute": 22.678, "trainingCostUsd2023": 266200.0315979732, "powerDrawW": 939481.3420790087, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2006.16668" }, { "name": "DeLighT", "organization": "University of Washington,Allen Institute for AI,Facebook AI Research", "orgCategory": "Academia,Research collective,Industry", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2020-08-03", "year": 2020, "decimalYear": 2020.587432, "domains": [ "Language" ], "task": "Language modeling,Translation", "parameters": 99000000, "computeFlop": 3801600000000000000, "log10Compute": 18.58, "trainingCostUsd2023": null, "powerDrawW": 4889.428515149248, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2008.00623" }, { "name": "ERNIE-GEN (large)", "organization": "Baidu", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2020-08-06", "year": 2020, "decimalYear": 2020.595628, "domains": [ "Language" ], "task": "Language modeling,Language modeling/generation,Text summarization,Chat", "parameters": 340000000, "computeFlop": 200000000000000000000, "log10Compute": 20.301, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Speculative", "link": "https://arxiv.org/abs/2001.11314" }, { "name": "ProBERTa", "organization": "University of Illinois Urbana-Champaign (UIUC),Reed College", "orgCategory": "Academia,Academia", "country": "United States", "countries": [ "United States" ], "date": "2020-09-01", "year": 2020, "decimalYear": 2020.666667, "domains": [ "Biology" ], "task": "Proteins,Protein representation learning,Protein classification,Protein interaction prediction", "parameters": 44000000, "computeFlop": 9720000000000000000, "log10Compute": 18.988, "trainingCostUsd2023": 26.206992435694062, "powerDrawW": 2443.1359420423914, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://dl.acm.org/doi/10.1145/3388440.3412467" }, { "name": "LUKE", "organization": "University of Washington,National Institute of Informatics", "orgCategory": "Academia", "country": "United States, Japan", "countries": [ "United States", "Japan" ], "date": "2020-10-02", "year": 2020, "decimalYear": 2020.751366, "domains": [ "Language" ], "task": "Question answering,Relation extraction,Named entity recognition (NER)", "parameters": 483000000, "computeFlop": 1.8144e+22, "log10Compute": 22.259, "trainingCostUsd2023": 4186.383565732907, "powerDrawW": 9765.799615782105, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2010.01057v1" }, { "name": "mT5-XXL", "organization": "Google,Google Research", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-10-20", "year": 2020, "decimalYear": 2020.800546, "domains": [ "Language" ], "task": "Language modeling,Translation,Language modeling/generation,Question answering", "parameters": 13000000000, "computeFlop": 8.2e+22, "log10Compute": 22.914, "trainingCostUsd2023": 179442.0291399545, "powerDrawW": null, "frontier": true, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://aclanthology.org/2021.naacl-main.41/" }, { "name": "Conformer + Wav2vec 2.0 + Noisy Student", "organization": "Google,Google Research,Google Brain", "orgCategory": "Industry,Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-10-20", "year": 2020, "decimalYear": 2020.800546, "domains": [ "Speech" ], "task": "Speech recognition (ASR)", "parameters": 1000000000, "computeFlop": 7.6e+21, "log10Compute": 21.881, "trainingCostUsd2023": 9449.541661137231, "powerDrawW": 234285.25901630375, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2010.10504v2" }, { "name": "German ELECTRA Large", "organization": "deepset,Bayerische Staatsbibliothek Muenchen", "orgCategory": "Industry,Government", "country": "Germany", "countries": [ "Germany" ], "date": "2020-10-21", "year": 2020, "decimalYear": 2020.803279, "domains": [ "Language" ], "task": "Document classification,Named entity recognition (NER),Text classification", "parameters": 335000000, "computeFlop": 1.42829568e+21, "log10Compute": 21.155, "trainingCostUsd2023": 2392.2883349806307, "powerDrawW": 58570.010422244966, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2010.10906" }, { "name": "wave2vec 2.0 LARGE", "organization": "Facebook", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-10-22", "year": 2020, "decimalYear": 2020.806011, "domains": [ "Speech" ], "task": "Speech completion", "parameters": 317000000, "computeFlop": 3.87072e+21, "log10Compute": 21.588, "trainingCostUsd2023": 5021.241075362514, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2006.11477" }, { "name": "ViT-Huge/14", "organization": "Google Brain,Google Research", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2020-10-22", "year": 2020, "decimalYear": 2020.806011, "domains": [ "Vision" ], "task": "Image representation", "parameters": 632000000, "computeFlop": 4.262e+21, "log10Compute": 21.63, "trainingCostUsd2023": 6724.023241261178, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2010.11929" }, { "name": "KEPLER", "organization": "Tsinghua University,Mila - Quebec AI (originally Montreal Institute for Learning Algorithms),HEC,CIFAR AI Research,Princeton University,University of Montreal / Université de Montréal", "orgCategory": "Academia,Academia,Academia,Research collective,Academia,Academia", "country": "China, Canada, France, United States", "countries": [ "China", "Canada", "France", "United States" ], "date": "2020-11-23", "year": 2020, "decimalYear": 2020.893443, "domains": [ "Language" ], "task": "Relation extraction", "parameters": 125000000, "computeFlop": 1.66e+21, "log10Compute": 21.22, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/1911.06136" }, { "name": "AlphaFold 2", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2020-11-30", "year": 2020, "decimalYear": 2020.912568, "domains": [ "Biology" ], "task": "Protein folding prediction,Proteins", "parameters": 93000000, "computeFlop": 2.99e+21, "log10Compute": 21.476, "trainingCostUsd2023": 3841.772612266474, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://www.nature.com/articles/s41586-021-03819-2" }, { "name": "CPM-Large", "organization": "Tsinghua University,Beijing Academy of Artificial Intelligence / BAAI", "orgCategory": "Academia,Academia", "country": "China", "countries": [ "China" ], "date": "2020-12-01", "year": 2020, "decimalYear": 2020.915301, "domains": [ "Language" ], "task": "Language modeling", "parameters": 2600000000, "computeFlop": 260520000000000000000, "log10Compute": 20.416, "trainingCostUsd2023": 7340.099044719796, "powerDrawW": 39011.038545758856, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2012.00413" }, { "name": "ESM1b", "organization": "Facebook AI Research,New York University (NYU)", "orgCategory": "Industry,Academia", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2020-12-15", "year": 2020, "decimalYear": 2020.953552, "domains": [ "Biology" ], "task": "Proteins,Protein or nucleotide language model (pLM/nLM)", "parameters": 652400000, "computeFlop": 5.1e+21, "log10Compute": 21.708, "trainingCostUsd2023": 924.3248891034536, "powerDrawW": 77997.75584661258, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://www.pnas.org/doi/abs/10.1073/pnas.2016239118" }, { "name": "DensePhrases", "organization": "Korea University,Princeton University", "orgCategory": "Academia,Academia", "country": "South Korea, United States", "countries": [ "South Korea", "United States" ], "date": "2020-12-23", "year": 2020, "decimalYear": 2020.97541, "domains": [ "Language" ], "task": "Question answering", "parameters": null, "computeFlop": 2099520000000000000, "log10Compute": 18.322, "trainingCostUsd2023": null, "powerDrawW": 4061.6594477324606, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Speculative", "link": "https://arxiv.org/abs/2012.12624v3" }, { "name": "CT-MoS (WT2)", "organization": "Google,National Tsing Hua University", "orgCategory": "Industry,Academia", "country": "United States, Taiwan", "countries": [ "United States", "Taiwan" ], "date": "2020-12-25", "year": 2020, "decimalYear": 2020.980874, "domains": [ "Language" ], "task": "Language modeling", "parameters": 45000000, "computeFlop": 540000000000000000, "log10Compute": 17.732, "trainingCostUsd2023": null, "powerDrawW": 2030.739275278243, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2012.13575" }, { "name": "ERNIE-Doc (247M)", "organization": "Baidu", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2020-12-31", "year": 2020, "decimalYear": 2020.997268, "domains": [ "Language" ], "task": "Language modeling,Language modeling/generation", "parameters": 247000000.00000003, "computeFlop": 30302798000000000000, "log10Compute": 19.481, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2012.15688" }, { "name": "DALL-E", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-01-05", "year": 2021, "decimalYear": 2021.010959, "domains": [ "Image generation" ], "task": "Text-to-image,Image generation", "parameters": 12000000000, "computeFlop": 4.7e+22, "log10Compute": 22.672, "trainingCostUsd2023": 125818.6901, "powerDrawW": 519741.92129196384, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Likely", "link": "https://openai.com/blog/dall-e/\n\nhttps://arxiv.org/abs/2102.12092" }, { "name": "CLIP (ViT L/14@336px)", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-01-05", "year": 2021, "decimalYear": 2021.010959, "domains": [ "Multimodal", "Vision", "Language", "Video" ], "task": "Zero-shot image classification,Character recognition (OCR),Video description", "parameters": 370000000, "computeFlop": 1.05e+22, "log10Compute": 22.021, "trainingCostUsd2023": 24638.518413409514, "powerDrawW": 155922.57638758916, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2103.00020\nhttps://huggingface.co/openai/clip-vit-large-patch14-336" }, { "name": "Switch", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-01-11", "year": 2021, "decimalYear": 2021.027397, "domains": [ "Language" ], "task": "Text autocompletion", "parameters": 1571000000000, "computeFlop": 8.22e+22, "log10Compute": 22.915, "trainingCostUsd2023": 145100.89767585773, "powerDrawW": 935410.4639575825, "frontier": true, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2101.03961" }, { "name": "DeiT-B", "organization": "Meta AI,Sorbonne University", "orgCategory": "Industry,Academia", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2021-01-15", "year": 2021, "decimalYear": 2021.038356, "domains": [ "Vision" ], "task": "Image classification", "parameters": 86000000, "computeFlop": 78840000000000000000, "log10Compute": 19.897, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2012.12877" }, { "name": "DLWP", "organization": "University of Washington,Microsoft Research", "orgCategory": "Academia,Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-02-09", "year": 2021, "decimalYear": 2021.106849, "domains": [ "Earth science" ], "task": "Weather forecasting", "parameters": 2676376, "computeFlop": 5684515200000000000, "log10Compute": 18.755, "trainingCostUsd2023": null, "powerDrawW": 278.6620974208019, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2102.05107" }, { "name": "MSA Transformer", "organization": "Facebook AI Research,University of California (UC) Berkeley,New York University (NYU)", "orgCategory": "Industry,Academia,Academia", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2021-02-13", "year": 2021, "decimalYear": 2021.117808, "domains": [ "Biology" ], "task": "Proteins,Protein or nucleotide language model (pLM/nLM),Protein contact and distance prediction,Protein folding prediction", "parameters": 100000000, "computeFlop": 5.49e+21, "log10Compute": 21.74, "trainingCostUsd2023": 13256.937301517895, "powerDrawW": 16227.834954676886, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://proceedings.mlr.press/v139/rao21a/rao21a.pdf" }, { "name": "SRU++ Large", "organization": "ASAPP", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-02-24", "year": 2021, "decimalYear": 2021.147945, "domains": [ "Language" ], "task": "Language modeling", "parameters": 234000000, "computeFlop": 21173704000000000000, "log10Compute": 19.326, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2102.12459" }, { "name": "Meta Pseudo Labels", "organization": "Google Brain,Google AI", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-03-01", "year": 2021, "decimalYear": 2021.161644, "domains": [ "Vision" ], "task": "Image classification", "parameters": 480000000, "computeFlop": 4.79e+22, "log10Compute": 22.68, "trainingCostUsd2023": 53844.28059190435, "powerDrawW": 934390.3015161053, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2003.10580" }, { "name": "M6-T", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2021-03-05", "year": 2021, "decimalYear": 2021.172603, "domains": [ "Multimodal", "Language", "Vision" ], "task": "Chat,Image captioning", "parameters": 1002700000000, "computeFlop": 5.5e+21, "log10Compute": 21.74, "trainingCostUsd2023": 13156.86123531961, "powerDrawW": 243309.1333554511, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2105.15082" }, { "name": "Generative BST", "organization": "Facebook AI Research", "orgCategory": "Industry", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2021-03-05", "year": 2021, "decimalYear": 2021.172603, "domains": [ "Language" ], "task": "Language modeling/generation,Chat,Question answering", "parameters": 9431810048, "computeFlop": 1.449e+22, "log10Compute": 22.161, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2004.13637" }, { "name": "PLUG", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2021-04-19", "year": 2021, "decimalYear": 2021.29589, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 27000000000, "computeFlop": 3.5997696e+22, "log10Compute": 22.556, "trainingCostUsd2023": 108672.69136370042, "powerDrawW": 103707.91685199105, "frontier": false, "openWeights": false, "accessibility": "Hosted access (no API)", "confidence": "Confident", "link": "https://mp.weixin.qq.com/s/DAQomIkDa52Sef-ruyH5qg" }, { "name": "ProtT5-XL-U50", "organization": "Technical University of Munich,Med AI Technology,NVIDIA,Oak Ridge National Laboratory,Google,Seoul National University", "orgCategory": "Academia,Industry,Government,Industry,Academia", "country": "Germany, China, United States, South Korea", "countries": [ "Germany", "China", "United States", "South Korea" ], "date": "2021-05-04", "year": 2021, "decimalYear": 2021.336986, "domains": [ "Biology" ], "task": "Proteins,Protein or nucleotide language model (pLM/nLM),Protein contact and distance prediction", "parameters": 3000000000, "computeFlop": 1.8704498688e+22, "log10Compute": 22.272, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://www.biorxiv.org/content/10.1101/2020.07.12.199554v3.full.pdf" }, { "name": "ProtBERT-BFD", "organization": "Technical University of Munich,NVIDIA,Seoul National University,Google,Oak Ridge National Laboratory,Med AI Technology", "orgCategory": "Academia,Industry,Academia,Industry,Government", "country": "Germany, United States, South Korea, China", "countries": [ "Germany", "United States", "South Korea", "China" ], "date": "2021-05-04", "year": 2021, "decimalYear": 2021.336986, "domains": [ "Biology" ], "task": "Proteins,Protein or nucleotide language model (pLM/nLM)", "parameters": 420000000, "computeFlop": 3.9e+22, "log10Compute": 22.591, "trainingCostUsd2023": 45350.735956744036, "powerDrawW": 933059.519872234, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://www.biorxiv.org/content/10.1101/2020.07.12.199554v3 or \nhttps://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9477085" }, { "name": "ADM", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-05-11", "year": 2021, "decimalYear": 2021.356164, "domains": [ "Image generation" ], "task": "Image generation,Text-to-image", "parameters": 559000000, "computeFlop": 6.2e+21, "log10Compute": 21.792, "trainingCostUsd2023": 11274.484326547095, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2105.05233" }, { "name": "MedBERT", "organization": "Peng Cheng Laboratory,University of Texas at Houston", "orgCategory": "Academia,Academia", "country": "China, United States", "countries": [ "China", "United States" ], "date": "2021-05-20", "year": 2021, "decimalYear": 2021.380822, "domains": [ "Medicine" ], "task": "Medical diagnosis,Text classification,Prediction of hospital stay duration,Prediction of diabetic heart failure (DHF),Prediction of onset of pancreatic cancer (PaCa)", "parameters": 17000000, "computeFlop": 9470000000000000000, "log10Compute": 18.976, "trainingCostUsd2023": 62.48645493610712, "powerDrawW": 278.0422249147422, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://www.nature.com/articles/s41746-021-00455-y" }, { "name": "Transformer local-attention (NesT-B)", "organization": "Google Cloud,Google Research", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-05-26", "year": 2021, "decimalYear": 2021.39726, "domains": [ "Vision" ], "task": "Image classification,Image generation", "parameters": 90100000, "computeFlop": 24057600000000000000, "log10Compute": 19.381, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2105.12723v4" }, { "name": "CogView", "organization": "Tsinghua University,Alibaba DAMO Academy", "orgCategory": "Academia,Industry", "country": "China", "countries": [ "China" ], "date": "2021-05-26", "year": 2021, "decimalYear": 2021.39726, "domains": [ "Image generation" ], "task": "Text-to-image,Image generation", "parameters": 4000000000, "computeFlop": 2.68e+22, "log10Compute": 22.428, "trainingCostUsd2023": 60071.706664791694, "powerDrawW": 259056.25043301377, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2105.13290" }, { "name": "ByT5-XXL", "organization": "Google,Google Research", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-05-28", "year": 2021, "decimalYear": 2021.40274, "domains": [ "Language" ], "task": "Language modeling", "parameters": 12900000000, "computeFlop": 8.1e+22, "log10Compute": 22.908, "trainingCostUsd2023": 92453.37635669747, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2105.13626" }, { "name": "ViT-G/14", "organization": "Google Brain,Google Research", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-06-08", "year": 2021, "decimalYear": 2021.432877, "domains": [ "Vision" ], "task": "Image classification", "parameters": 1843000000, "computeFlop": 5.85e+22, "log10Compute": 22.767, "trainingCostUsd2023": 3847.8613922131217, "powerDrawW": 1864665.1010174435, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2106.04560" }, { "name": "EMDR", "organization": "Mila - Quebec AI (originally Montreal Institute for Learning Algorithms),McGill University,DeepMind", "orgCategory": "Academia,Academia,Industry", "country": "Canada, United Kingdom", "countries": [ "Canada", "United Kingdom" ], "date": "2021-06-09", "year": 2021, "decimalYear": 2021.435616, "domains": [ "Language" ], "task": "Question answering", "parameters": 440000000, "computeFlop": 1.91e+21, "log10Compute": 21.281, "trainingCostUsd2023": 2773.3124307816734, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2106.05346v2" }, { "name": "CoAtNet", "organization": "Google,Google Research,Google Brain", "orgCategory": "Industry,Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-06-09", "year": 2021, "decimalYear": 2021.435616, "domains": [ "Vision" ], "task": "Image classification", "parameters": 2440000000, "computeFlop": 4.27e+22, "log10Compute": 22.63, "trainingCostUsd2023": 1887.1635925610951, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2106.04803v2" }, { "name": "DeBERTa", "organization": "Microsoft", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-06-10", "year": 2021, "decimalYear": 2021.438356, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering", "parameters": 1500000000, "computeFlop": 2.588e+22, "log10Compute": 22.413, "trainingCostUsd2023": 6682.2289986716, "powerDrawW": 155381.83775233384, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2006.03654" }, { "name": "Denoising Diffusion Probabilistic Models (LSUN Bedroom)", "organization": "University of California (UC) Berkeley", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2021-06-11", "year": 2021, "decimalYear": 2021.441096, "domains": [ "Vision" ], "task": "Image generation", "parameters": 256000000, "computeFlop": 78401250000000010000, "log10Compute": 19.894, "trainingCostUsd2023": 436.308484475363, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2006.11239" }, { "name": "ALIGN", "organization": "Google Research", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-06-11", "year": 2021, "decimalYear": 2021.441096, "domains": [ "Multimodal", "Vision", "Language" ], "task": "Representation learning,Image classification,Image representation", "parameters": 820000000, "computeFlop": 2.598670000001e+22, "log10Compute": 22.415, "trainingCostUsd2023": 32852.91660437908, "powerDrawW": 466135.13260508416, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2102.05918" }, { "name": "StyleGAN3-T", "organization": "NVIDIA,Aalto University", "orgCategory": "Industry,Academia", "country": "United States, Finland", "countries": [ "United States", "Finland" ], "date": "2021-06-21", "year": 2021, "decimalYear": 2021.468493, "domains": [ "Image generation" ], "task": "Image generation", "parameters": 2230000, "computeFlop": 1.70208e+21, "log10Compute": 21.231, "trainingCostUsd2023": null, "powerDrawW": 4854.493112492733, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2106.12423" }, { "name": "StyleGAN3-R", "organization": "NVIDIA,Aalto University", "orgCategory": "Industry,Academia", "country": "United States, Finland", "countries": [ "United States", "Finland" ], "date": "2021-06-21", "year": 2021, "decimalYear": 2021.468493, "domains": [ "Image generation" ], "task": "Image generation", "parameters": 1580000, "computeFlop": 2.42784e+21, "log10Compute": 21.385, "trainingCostUsd2023": null, "powerDrawW": 4854.493112492733, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/pdf/2106.12423" }, { "name": "EfficientNetV2-XL", "organization": "Google,Google Brain", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-06-23", "year": 2021, "decimalYear": 2021.473973, "domains": [ "Vision" ], "task": "Image classification,Neural Architecture Search - NAS", "parameters": 208000000, "computeFlop": 95600000000000000000, "log10Compute": 19.98, "trainingCostUsd2023": 104.34013401561589, "powerDrawW": 14562.830712865913, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2104.00298" }, { "name": "Fold2Seq", "organization": "IBM,Texas A&M", "orgCategory": "Industry,Academia", "country": "United States", "countries": [ "United States" ], "date": "2021-06-24", "year": 2021, "decimalYear": 2021.476712, "domains": [ "Biology" ], "task": "Proteins,Protein generation,Protein inverse folding,Protein fold classification", "parameters": 12427904, "computeFlop": 140000000000000000, "log10Compute": 17.146, "trainingCostUsd2023": null, "powerDrawW": 1213.542200949413, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://arxiv.org/abs/2106.13058" }, { "name": "Adaptive Input Transformer + RD", "organization": "Microsoft Research Asia,Soochow University", "orgCategory": "Industry,Academia", "country": "China, Taiwan", "countries": [ "China", "Taiwan" ], "date": "2021-06-28", "year": 2021, "decimalYear": 2021.487671, "domains": [ "Language" ], "task": "Language modeling", "parameters": 247000000.00000003, "computeFlop": 85800000000000000000, "log10Compute": 19.933, "trainingCostUsd2023": null, "powerDrawW": 4853.736425906205, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2106.14448" }, { "name": "ERNIE 3.0", "organization": "Baidu", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2021-07-05", "year": 2021, "decimalYear": 2021.506849, "domains": [ "Language" ], "task": "Language modeling,Language modeling/generation,Text classification,Question answering", "parameters": 10000000000, "computeFlop": 2.25e+22, "log10Compute": 22.352, "trainingCostUsd2023": 39104.26710360624, "powerDrawW": 232943.03314881725, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2107.02137" }, { "name": "Codex", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-07-07", "year": 2021, "decimalYear": 2021.512329, "domains": [ "Language" ], "task": "Code autocompletion", "parameters": 12000000000, "computeFlop": 7.344e+22, "log10Compute": 22.866, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Likely", "link": "https://openai.com/blog/openai-codex/\nhttps://arxiv.org/abs/2107.03374" }, { "name": "HuBERT", "organization": "Facebook AI Research", "orgCategory": "Industry", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2021-07-27", "year": 2021, "decimalYear": 2021.567123, "domains": [ "Speech" ], "task": "Speech recognition (ASR)", "parameters": 1000000000, "computeFlop": 5.54e+21, "log10Compute": 21.744, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Speculative", "link": "https://arxiv.org/abs/2106.07447" }, { "name": "GOAT", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2021-07-27", "year": 2021, "decimalYear": 2021.567123, "domains": [ "Games" ], "task": "Open ended play", "parameters": 3472816, "computeFlop": 2.412e+22, "log10Compute": 22.382, "trainingCostUsd2023": 84799.78517435073, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://deepmind.com/blog/article/generally-capable-agents-emerge-from-open-ended-play\n\nhttps://arxiv.org/abs/2107.12808" }, { "name": "SEER", "organization": "Facebook AI Research,INRIA", "orgCategory": "Industry,Academia", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2021-07-29", "year": 2021, "decimalYear": 2021.572603, "domains": [ "Vision" ], "task": "Image embedding,Image classification", "parameters": 1300000000, "computeFlop": 1.8e+22, "log10Compute": 22.255, "trainingCostUsd2023": 34114.252463690456, "powerDrawW": 310424.75538121304, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2103.01988" }, { "name": "YOLOX-X", "organization": "Megvii Inc", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2021-08-06", "year": 2021, "decimalYear": 2021.594521, "domains": [ "Vision" ], "task": "Object detection", "parameters": 99100000, "computeFlop": 634275000000000000000, "log10Compute": 20.802, "trainingCostUsd2023": null, "powerDrawW": 4849.522759284234, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2107.08430" }, { "name": "Zidong Taichu", "organization": "Chinese Academy of Sciences,Wuhan AI Computing Center", "orgCategory": "Academia,Government", "country": "China", "countries": [ "China" ], "date": "2021-08-11", "year": 2021, "decimalYear": 2021.608219, "domains": [ "Multimodal", "Speech", "Vision", "Language" ], "task": "Language modeling/generation,Speech recognition (ASR),Image captioning", "parameters": 3200000000, "computeFlop": 801600000000000000000, "log10Compute": 20.904, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://gitee.com/zidongtaichu/multi-modal-models" }, { "name": "Jurassic-1-Jumbo", "organization": "AI21 Labs", "orgCategory": "Industry", "country": "Israel", "countries": [ "Israel" ], "date": "2021-08-11", "year": 2021, "decimalYear": 2021.608219, "domains": [ "Language" ], "task": "Language modeling/generation,Chat", "parameters": 178000000000, "computeFlop": 3.7e+23, "log10Compute": 23.568, "trainingCostUsd2023": 836700.050829598, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Confident", "link": "https://uploads-ssl.webflow.com/60fd4503684b466578c0d307/61138924626a6981ee09caf6_jurassic_tech_paper.pdf" }, { "name": "DNABERT", "organization": "Northeastern University", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2021-08-15", "year": 2021, "decimalYear": 2021.619178, "domains": [ "Biology" ], "task": "Protein or nucleotide language model (pLM/nLM)", "parameters": 110000000, "computeFlop": 107000000000000000000, "log10Compute": 20.029, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://academic.oup.com/bioinformatics/article/37/15/2112/6128680" }, { "name": "XLMR-XXL", "organization": "Facebook AI Research", "orgCategory": "Industry", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2021-08-17", "year": 2021, "decimalYear": 2021.624658, "domains": [ "Language" ], "task": "Translation,Language modeling/generation", "parameters": 10700000000, "computeFlop": 3.366e+22, "log10Compute": 22.527, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2105.00572" }, { "name": "FLAN 137B", "organization": "Google Research", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-09-03", "year": 2021, "decimalYear": 2021.671233, "domains": [ "Language" ], "task": "Language modeling,Question answering,Language modeling/generation", "parameters": 137000000000, "computeFlop": 2.047e+24, "log10Compute": 24.311, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2109.01652" }, { "name": "PermuteFormer", "organization": "Peking University", "orgCategory": "Academia", "country": "China", "countries": [ "China" ], "date": "2021-09-06", "year": 2021, "decimalYear": 2021.679452, "domains": [ "Language" ], "task": "Language modeling", "parameters": 149697024, "computeFlop": 2775000000000000000, "log10Compute": 18.443, "trainingCostUsd2023": null, "powerDrawW": 4846.17604411253, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://arxiv.org/abs/2109.02377" }, { "name": "HyperCLOVA 204B", "organization": "NAVER", "orgCategory": "Industry", "country": "South Korea", "countries": [ "South Korea" ], "date": "2021-09-10", "year": 2021, "decimalYear": 2021.690411, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 204000000000, "computeFlop": 2.0000000001e+23, "log10Compute": 23.301, "trainingCostUsd2023": 441770.0038182093, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "Hosted access (no API)", "confidence": "Speculative", "link": "https://aibusiness.com/nlp/south-korea-s-naver-unveils-hyperscale-ai-platform-language-model-with-more-parameters-than-gpt-3" }, { "name": "PLATO-XL", "organization": "Baidu", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2021-09-20", "year": 2021, "decimalYear": 2021.717808, "domains": [ "Language" ], "task": "Language modeling", "parameters": 11000000000, "computeFlop": 9.9e+21, "log10Compute": 21.996, "trainingCostUsd2023": null, "powerDrawW": 155029.29220816068, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2109.09519" }, { "name": "Turing ULRv5", "organization": "Microsoft", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-09-28", "year": 2021, "decimalYear": 2021.739726, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Translation", "parameters": 2200000000, "computeFlop": 2.8983951e+22, "log10Compute": 22.462, "trainingCostUsd2023": null, "powerDrawW": 206668.900572444, "frontier": false, "openWeights": false, "accessibility": "Hosted access (no API)", "confidence": "Confident", "link": "https://www.microsoft.com/en-us/research/blog/microsoft-turing-universal-language-representation-model-t-ulrv5-tops-xtreme-leaderboard-and-trains-100x-faster/" }, { "name": "AlphaFold-Multimer", "organization": "Google DeepMind,DeepMind", "orgCategory": "Industry,Industry", "country": "United States, United Kingdom", "countries": [ "United States", "United Kingdom" ], "date": "2021-10-04", "year": 2021, "decimalYear": 2021.756164, "domains": [ "Biology" ], "task": "Protein folding prediction,Proteins", "parameters": null, "computeFlop": 4.35e+21, "log10Compute": 21.638, "trainingCostUsd2023": 7966.2330131223, "powerDrawW": 58117.86227766231, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://www.biorxiv.org/content/10.1101/2021.10.04.463034v1" }, { "name": "Megatron-Turing NLG 530B", "organization": "Microsoft,NVIDIA", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-10-11", "year": 2021, "decimalYear": 2021.775342, "domains": [ "Language" ], "task": "Language modeling,Language modeling/generation,Question answering", "parameters": 530000000000, "computeFlop": 8.586e+23, "log10Compute": 23.934, "trainingCostUsd2023": 3848505.6256471733, "powerDrawW": 3615658.868639838, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2201.11990" }, { "name": "Yuan 1.0", "organization": "Inspur", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2021-10-12", "year": 2021, "decimalYear": 2021.778082, "domains": [ "Language" ], "task": "Language modeling", "parameters": 245730000000, "computeFlop": 3.5380000000001e+23, "log10Compute": 23.549, "trainingCostUsd2023": 591664.5798708292, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Confident", "link": "https://arxiv.org/abs/2110.04725" }, { "name": "base LM+GNN+kNN", "organization": "Shannon.AI,Nanjing University,Nanyang Technological University,Zhejiang University (ZJU)", "orgCategory": "Industry,Academia,Academia,Academia", "country": "China, Singapore", "countries": [ "China", "Singapore" ], "date": "2021-10-17", "year": 2021, "decimalYear": 2021.791781, "domains": [ "Language" ], "task": "Language modeling", "parameters": 274000000, "computeFlop": 52587456000000000000, "log10Compute": 19.721, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2110.08743" }, { "name": "S4", "organization": "Stanford University", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2021-10-31", "year": 2021, "decimalYear": 2021.830137, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 249000000.00000003, "computeFlop": 78328627000000000000, "log10Compute": 19.894, "trainingCostUsd2023": null, "powerDrawW": 6453.658675376135, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2111.00396" }, { "name": "Projected GAN", "organization": "Heidelberg University", "orgCategory": "Academia", "country": "Germany", "countries": [ "Germany" ], "date": "2021-11-01", "year": 2021, "decimalYear": 2021.832877, "domains": [ "Image generation" ], "task": "Image generation", "parameters": null, "computeFlop": 10500000000000000000, "log10Compute": 19.021, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://proceedings.neurips.cc/paper/2021/hash/9219adc5c42107c4911e249155320648-Abstract.html" }, { "name": "CodeT5-base", "organization": "Salesforce,Nanyang Technological University", "orgCategory": "Industry,Academia", "country": "United States, Singapore", "countries": [ "United States", "Singapore" ], "date": "2021-11-01", "year": 2021, "decimalYear": 2021.832877, "domains": [ "Language" ], "task": "Code generation", "parameters": 220000000, "computeFlop": 1.56e+21, "log10Compute": 21.193, "trainingCostUsd2023": 3114.8690946174474, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://aclanthology.org/2021.emnlp-main.685/" }, { "name": "Masked Autoencoders ViT-H", "organization": "Facebook AI Research", "orgCategory": "Industry", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2021-11-11", "year": 2021, "decimalYear": 2021.860274, "domains": [ "Vision" ], "task": "Semantic segmentation,Image classification,Image generation", "parameters": 632000000, "computeFlop": 460000000000000000000, "log10Compute": 20.663, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Speculative", "link": "https://arxiv.org/abs/2111.06377" }, { "name": "Swin Transformer V2 (SwinV2-G)", "organization": "Microsoft Research Asia", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2021-11-18", "year": 2021, "decimalYear": 2021.879452, "domains": [ "Vision", "Video" ], "task": "Action recognition,Image classification", "parameters": 3000000000, "computeFlop": 1.1e+21, "log10Compute": 21.041, "trainingCostUsd2023": 2326.6636503781665, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2111.09883v2" }, { "name": "BASIC-L", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-11-19", "year": 2021, "decimalYear": 2021.882192, "domains": [ "Vision" ], "task": "Image classification", "parameters": 3070000000, "computeFlop": 4.12e+22, "log10Compute": 22.615, "trainingCostUsd2023": 1684.770712126102, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2111.10050" }, { "name": "Florence", "organization": "Microsoft", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-11-22", "year": 2021, "decimalYear": 2021.890411, "domains": [ "Vision" ], "task": "Image captioning,Visual question answering,Image classification,Object detection", "parameters": 893000000, "computeFlop": 4.831e+22, "log10Compute": 22.684, "trainingCostUsd2023": 106950.61569328007, "powerDrawW": 412831.84854484117, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2111.11432v1" }, { "name": "NÜWA", "organization": "Microsoft Research,Peking University", "orgCategory": "Industry,Academia", "country": "United States, China", "countries": [ "United States", "China" ], "date": "2021-11-24", "year": 2021, "decimalYear": 2021.89589, "domains": [ "Multimodal", "Vision", "Image generation", "Video", "Language" ], "task": "Image generation,Video generation,Text-to-image,Text-to-video", "parameters": 870000000, "computeFlop": 7.24598784e+21, "log10Compute": 21.86, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2111.12417" }, { "name": "Student of Games", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2021-12-06", "year": 2021, "decimalYear": 2021.928767, "domains": [ "Games" ], "task": "Chess,Go,Poker", "parameters": null, "computeFlop": 3.6679273004682866e+22, "log10Compute": 22.564, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://arxiv.org/abs/2112.03178" }, { "name": "Gopher (280B)", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2021-12-08", "year": 2021, "decimalYear": 2021.934247, "domains": [ "Language" ], "task": "Language modeling,Question answering", "parameters": 280000000000, "computeFlop": 6.31e+23, "log10Compute": 23.8, "trainingCostUsd2023": 640616.8522346151, "powerDrawW": 3714163.0079064844, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2112.11446" }, { "name": "GLaM", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2021-12-13", "year": 2021, "decimalYear": 2021.947945, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering", "parameters": 1200000000000, "computeFlop": 3.6363112434e+23, "log10Compute": 23.561, "trainingCostUsd2023": 541437.4162400038, "powerDrawW": 701486.0111048243, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2112.06905" }, { "name": "Contriever", "organization": "Meta AI,University College London (UCL),PSL University,Université Grenoble Alpes", "orgCategory": "Industry,Academia,Academia,Academia", "country": "United States, United Kingdom, France", "countries": [ "United States", "United Kingdom", "France" ], "date": "2021-12-16", "year": 2021, "decimalYear": 2021.956164, "domains": [ "Language" ], "task": "Language modeling", "parameters": 110000000, "computeFlop": 157000000000000000000, "log10Compute": 20.196, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Likely", "link": "https://arxiv.org/abs/2112.09118" }, { "name": "XGLM-7.5B", "organization": "Meta AI,Facebook AI Research", "orgCategory": "Industry,Industry", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2021-12-20", "year": 2021, "decimalYear": 2021.967123, "domains": [ "Language" ], "task": "Translation,Question answering,Language modeling/generation", "parameters": 7500000000, "computeFlop": 2.25e+22, "log10Compute": 22.352, "trainingCostUsd2023": 104152.22590136188, "powerDrawW": 206287.25531193183, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2112.10668" }, { "name": "ERNIE 3.0 Titan", "organization": "Baidu,Peng Cheng Laboratory", "orgCategory": "Industry,Academia", "country": "China", "countries": [ "China" ], "date": "2021-12-23", "year": 2021, "decimalYear": 2021.975342, "domains": [ "Language" ], "task": "Language modeling,Language modeling/generation,Relation extraction,Sentiment classification,Text classification", "parameters": 260000000000, "computeFlop": 1.0421e+24, "log10Compute": 24.018, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "Hosted access (no API)", "confidence": "Confident", "link": "https://arxiv.org/abs/2112.12731" }, { "name": "Detic", "organization": "Meta AI,University of Texas at Austin", "orgCategory": "Industry,Academia", "country": "United States", "countries": [ "United States" ], "date": "2022-01-07", "year": 2022, "decimalYear": 2022.016438, "domains": [ "Vision" ], "task": "Object detection,Image classification", "parameters": 88000000, "computeFlop": 23439974400000000000, "log10Compute": 19.37, "trainingCostUsd2023": 191.44581825616135, "powerDrawW": 19331.67955389363, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Speculative", "link": "https://arxiv.org/abs/2201.02605" }, { "name": "InstructGPT 175B", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-01-27", "year": 2022, "decimalYear": 2022.071233, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 175000000000, "computeFlop": 3.19181e+23, "log10Compute": 23.504, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Confident", "link": "https://arxiv.org/pdf/2203.02155" }, { "name": "AlphaCode", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2022-02-02", "year": 2022, "decimalYear": 2022.087671, "domains": [ "Language" ], "task": "Code generation", "parameters": 41100000000, "computeFlop": 2.38010000000001e+23, "log10Compute": 23.377, "trainingCostUsd2023": null, "powerDrawW": 2566002.5357062262, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2203.07814" }, { "name": "RETRO-7B", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2022-02-07", "year": 2022, "decimalYear": 2022.10137, "domains": [ "Language" ], "task": "Language modeling/generation,Language modeling", "parameters": 7500000000, "computeFlop": 1.68e+22, "log10Compute": 22.225, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2112.04426" }, { "name": "GPT-NeoX-20B", "organization": "EleutherAI", "orgCategory": "Research collective", "country": "United States", "countries": [ "United States" ], "date": "2022-02-09", "year": 2022, "decimalYear": 2022.106849, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 20000000000, "computeFlop": 9.31627008e+22, "log10Compute": 22.969, "trainingCostUsd2023": 184272.80736002637, "powerDrawW": 77269.91251696396, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2204.06745" }, { "name": "ProteinBERT", "organization": "Hebrew University of Jerusalem,Ben-Gurion University of the Negev,Deep Trading", "orgCategory": "Academia,Academia,Industry", "country": "Israel, United States", "countries": [ "Israel", "United States" ], "date": "2022-02-10", "year": 2022, "decimalYear": 2022.109589, "domains": [ "Biology" ], "task": "Proteins,Protein generation,Protein representation learning", "parameters": 16000000, "computeFlop": 65000000000000000000, "log10Compute": 19.813, "trainingCostUsd2023": null, "powerDrawW": 254.28806247515058, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://academic.oup.com/bioinformatics/article/38/8/2102/6502274" }, { "name": "LaMDA", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-02-10", "year": 2022, "decimalYear": 2022.109589, "domains": [ "Language" ], "task": "Language modeling", "parameters": 137000000000, "computeFlop": 3.55e+23, "log10Compute": 23.55, "trainingCostUsd2023": 229949.98625999544, "powerDrawW": 927218.301405043, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2201.08239" }, { "name": "ST-MoE", "organization": "Google,Google Brain,Google Research", "orgCategory": "Industry,Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-02-17", "year": 2022, "decimalYear": 2022.128767, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 269000000000, "computeFlop": 2.9e+23, "log10Compute": 23.462, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2202.08906v2" }, { "name": "FourCastNet", "organization": "NVIDIA,NERSC, Lawrence Berkeley National Laboratory,University of Michigan,Rice University,California Institute of Technology,Purdue University", "orgCategory": "Industry,Government,Academia,Academia,Academia,Academia", "country": "United States", "countries": [ "United States" ], "date": "2022-02-22", "year": 2022, "decimalYear": 2022.142466, "domains": [ "Earth science" ], "task": "Weather forecasting", "parameters": null, "computeFlop": 345047040000000000000, "log10Compute": 20.538, "trainingCostUsd2023": null, "powerDrawW": 51498.36398258959, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2202.11214" }, { "name": "PolyCoder", "organization": "Carnegie Mellon University (CMU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2022-02-26", "year": 2022, "decimalYear": 2022.153425, "domains": [ "Language" ], "task": "Code generation", "parameters": 2700000000, "computeFlop": 1.1e+21, "log10Compute": 21.041, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2202.13169" }, { "name": "Statement Curriculum Learning", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-03-02", "year": 2022, "decimalYear": 2022.164384, "domains": [ "Language", "Mathematics" ], "task": "Automated theorem proving,Language modeling/generation,Question answering,Mathematical reasoning", "parameters": 774000000, "computeFlop": 1.7901648e+22, "log10Compute": 22.253, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Likely", "link": "https://arxiv.org/abs/2202.01344" }, { "name": "ViT-G (model soup)", "organization": "University of Washington,Columbia University,Google,Meta AI,Tel Aviv University", "orgCategory": "Academia,Academia,Industry,Industry,Academia", "country": "United States, Israel", "countries": [ "United States", "Israel" ], "date": "2022-03-10", "year": 2022, "decimalYear": 2022.186301, "domains": [ "Vision" ], "task": "Image classification", "parameters": 1843000000, "computeFlop": 3.4e+21, "log10Compute": 21.531, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2203.05482v3" }, { "name": "GPT-3.5 (davinci-002)", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-03-15", "year": 2022, "decimalYear": 2022.2, "domains": [ "Language" ], "task": "Language modeling", "parameters": null, "computeFlop": 2.578e+24, "log10Compute": 24.411, "trainingCostUsd2023": 4897551.634095244, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Speculative", "link": "https://platform.openai.com/docs/models/gpt-3-5" }, { "name": "Segatron-XL large, M=384 + HCP", "organization": "Microsoft Research,University of Waterloo", "orgCategory": "Industry,Academia", "country": "United States, Canada", "countries": [ "United States", "Canada" ], "date": "2022-03-21", "year": 2022, "decimalYear": 2022.216438, "domains": [ "Language" ], "task": "Language modeling", "parameters": 256999999.99999997, "computeFlop": 26500000000000000000, "log10Compute": 19.423, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2203.10692" }, { "name": "Make-A-Scene", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-03-24", "year": 2022, "decimalYear": 2022.224658, "domains": [ "Image generation" ], "task": "Image generation,Text-to-image", "parameters": 4000000000, "computeFlop": 6.4172851e+21, "log10Compute": 21.807, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2203.13131" }, { "name": "Chinchilla", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2022-03-29", "year": 2022, "decimalYear": 2022.238356, "domains": [ "Language" ], "task": "Language modeling", "parameters": 70000000000, "computeFlop": 5.76e+23, "log10Compute": 23.76, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2203.15556" }, { "name": "PaLM (540B)", "organization": "Google Research", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-04-04", "year": 2022, "decimalYear": 2022.254795, "domains": [ "Language" ], "task": "Language modeling,Code generation,Translation", "parameters": 540350000000, "computeFlop": 2.5272e+24, "log10Compute": 24.403, "trainingCostUsd2023": 3060364.5969860666, "powerDrawW": 4198431.396906301, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2204.02311" }, { "name": "DALL·E 2", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-04-06", "year": 2022, "decimalYear": 2022.260274, "domains": [ "Image generation" ], "task": "Text-to-image,Image generation", "parameters": 3500000000, "computeFlop": 3.3695784e+23, "log10Compute": 23.528, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Speculative", "link": "https://cdn.openai.com/papers/dall-e-2.pdf" }, { "name": "BERT-RBP", "organization": "Waseda University", "orgCategory": "Academia", "country": "Japan", "countries": [ "Japan" ], "date": "2022-04-07", "year": 2022, "decimalYear": 2022.263014, "domains": [ "Biology" ], "task": "Proteins,Protein interaction prediction,RNA-Protein interaction prediction", "parameters": 110000000, "computeFlop": 140000000000000000000, "log10Compute": 20.146, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://academic.oup.com/bioinformaticsadvances/article/2/1/vbac023/6564689" }, { "name": "Stable Diffusion (LDM-KL-8-G)", "organization": "Runway,Ludwig Maximilian University of Munich,Heidelberg University", "orgCategory": "Industry,Academia,Academia", "country": "United States, Germany", "countries": [ "United States", "Germany" ], "date": "2022-04-13", "year": 2022, "decimalYear": 2022.279452, "domains": [ "Image generation" ], "task": "Image generation,Text-to-image", "parameters": 1450000000, "computeFlop": 5.0000000000000004e+22, "log10Compute": 22.699, "trainingCostUsd2023": 111248.21698072631, "powerDrawW": 205764.21634205984, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2112.10752" }, { "name": "Sparse all-MLP", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-04-14", "year": 2022, "decimalYear": 2022.282192, "domains": [ "Language" ], "task": "Language modeling", "parameters": 9410000000, "computeFlop": 532224000000000000000, "log10Compute": 20.726, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2203.06850" }, { "name": "Flamingo", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2022-04-29", "year": 2022, "decimalYear": 2022.323288, "domains": [ "Multimodal", "Vision", "Language", "Video" ], "task": "Visual question answering,Image captioning", "parameters": 80000000000, "computeFlop": 2.18972000000001e+23, "log10Compute": 23.34, "trainingCostUsd2023": 183423.16330597224, "powerDrawW": 1049023.659188776, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2204.14198" }, { "name": "OPT-175B", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-05-02", "year": 2022, "decimalYear": 2022.331507, "domains": [ "Language" ], "task": "Language modeling,Chat,Language modeling/generation,Question answering", "parameters": 175000000000, "computeFlop": 4.3e+23, "log10Compute": 23.633, "trainingCostUsd2023": 733634.637, "powerDrawW": 822708.6888139298, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2205.01068" }, { "name": "UL2", "organization": "Google Research,Google Brain", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-05-10", "year": 2022, "decimalYear": 2022.353425, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Text summarization", "parameters": 20000000000, "computeFlop": 1.2e+23, "log10Compute": 23.079, "trainingCostUsd2023": 126785.76203549477, "powerDrawW": 349588.9061965891, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2205.05131v1" }, { "name": "Gato", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2022-05-12", "year": 2022, "decimalYear": 2022.358904, "domains": [ "Multimodal", "Robotics", "Games", "Language" ], "task": "Atari,Image captioning,Chat,Robotic manipulation", "parameters": 1180000000, "computeFlop": 4.02e+21, "log10Compute": 21.604, "trainingCostUsd2023": 3523.0649800752253, "powerDrawW": 231335.2960762711, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2205.06175" }, { "name": "Imagen", "organization": "Google Brain", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-05-23", "year": 2022, "decimalYear": 2022.389041, "domains": [ "Image generation" ], "task": "Text-to-image,Image generation", "parameters": 7762000000, "computeFlop": 1.46e+22, "log10Compute": 22.164, "trainingCostUsd2023": 7915.823806150154, "powerDrawW": 174743.8571090156, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Likely", "link": "https://arxiv.org/abs/2205.11487" }, { "name": "Tranception", "organization": "University of Oxford,Harvard Medical School,Cohere", "orgCategory": "Academia,Academia,Industry", "country": "United Kingdom, United States, Canada", "countries": [ "United Kingdom", "United States", "Canada" ], "date": "2022-05-27", "year": 2022, "decimalYear": 2022.4, "domains": [ "Biology" ], "task": "Proteins,Protein pathogenicity prediction", "parameters": 700000000, "computeFlop": 7.24e+21, "log10Compute": 21.86, "trainingCostUsd2023": 15247.43608848737, "powerDrawW": 51390.67413498394, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2205.13760" }, { "name": "GPT-2 Medium (FlashAttention)", "organization": "Stanford University,University at Buffalo", "orgCategory": "Academia,Academia", "country": "United States", "countries": [ "United States" ], "date": "2022-05-27", "year": 2022, "decimalYear": 2022.4, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 355000000, "computeFlop": 892809220000000000000, "log10Compute": 20.951, "trainingCostUsd2023": null, "powerDrawW": 6423.834266872993, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2205.14135" }, { "name": "DITTO", "organization": "Tsinghua University,Apple,Westlake University,Chinese University of Hong Kong (CUHK)", "orgCategory": "Academia,Industry,Academia,Academia", "country": "China, United States, Hong Kong", "countries": [ "China", "United States", "Hong Kong" ], "date": "2022-06-06", "year": 2022, "decimalYear": 2022.427397, "domains": [ "Language" ], "task": "Language modeling/generation,Text summarization", "parameters": 750000000, "computeFlop": 3317760000000000000, "log10Compute": 18.521, "trainingCostUsd2023": null, "powerDrawW": 4816.802908983849, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2206.02369" }, { "name": "CoCa", "organization": "Google Research", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-06-14", "year": 2022, "decimalYear": 2022.449315, "domains": [ "Vision" ], "task": "Image classification,Visual question answering,Image captioning", "parameters": 2100000000, "computeFlop": 7.3e+22, "log10Compute": 22.863, "trainingCostUsd2023": 78043.3756911775, "powerDrawW": 1397266.1319117176, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2205.01917v2" }, { "name": "Parti", "organization": "Google Research", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-06-22", "year": 2022, "decimalYear": 2022.471233, "domains": [ "Image generation" ], "task": "Text-to-image,Image generation", "parameters": 20000000000, "computeFlop": 5.09607936e+23, "log10Compute": 23.707, "trainingCostUsd2023": 427178.712, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2206.10789v1" }, { "name": "ProGen2-xlarge", "organization": "Salesforce Research,Columbia University,Johns Hopkins University", "orgCategory": "Industry,Academia,Academia", "country": "United States", "countries": [ "United States" ], "date": "2022-06-27", "year": 2022, "decimalYear": 2022.484932, "domains": [ "Biology" ], "task": "Proteins,Protein generation,Protein or nucleotide language model (pLM/nLM)", "parameters": 6400000000, "computeFlop": 1.35e+22, "log10Compute": 22.13, "trainingCostUsd2023": 11850.178410269727, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2206.13517" }, { "name": "Minerva (540B)", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-06-29", "year": 2022, "decimalYear": 2022.490411, "domains": [ "Language" ], "task": "Quantitative reasoning,Mathematical reasoning,Language modeling/generation,Question answering", "parameters": 540350000000, "computeFlop": 2.7415e+24, "log10Compute": 24.438, "trainingCostUsd2023": null, "powerDrawW": 698399.733142376, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2206.14858" }, { "name": "CodeT5-large", "organization": "Salesforce", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-07-05", "year": 2022, "decimalYear": 2022.506849, "domains": [ "Language" ], "task": "Code generation", "parameters": 770000000, "computeFlop": 2.72e+21, "log10Compute": 21.435, "trainingCostUsd2023": 4478.145684414144, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2207.01780" }, { "name": "NLLB", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-07-06", "year": 2022, "decimalYear": 2022.509589, "domains": [ "Language" ], "task": "Translation", "parameters": 54500000000, "computeFlop": 1.751113728e+22, "log10Compute": 22.243, "trainingCostUsd2023": 50667.25034038439, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://research.facebook.com/publications/no-language-left-behind/" }, { "name": "BLOOM-176B", "organization": "Hugging Face,BigScience", "orgCategory": "Industry,Research collective", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2022-07-11", "year": 2022, "decimalYear": 2022.523288, "domains": [ "Language" ], "task": "Language modeling,Translation,Code generation", "parameters": 176247271424, "computeFlop": 3.65664e+23, "log10Compute": 23.563, "trainingCostUsd2023": 995819.1171, "powerDrawW": 308035.2013244814, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://arxiv.org/abs/2211.05100" }, { "name": "ESM2-15B", "organization": "Meta AI,New York University (NYU),Stanford University,Massachusetts Institute of Technology (MIT)", "orgCategory": "Industry,Academia,Academia,Academia", "country": "United States", "countries": [ "United States" ], "date": "2022-07-21", "year": 2022, "decimalYear": 2022.550685, "domains": [ "Biology" ], "task": "Proteins,Protein or nucleotide language model (pLM/nLM),Protein folding prediction", "parameters": 15000000000, "computeFlop": 7.35000000001e+22, "log10Compute": 22.866, "trainingCostUsd2023": 163467.82019979745, "powerDrawW": 307966.61145938886, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://www.science.org/doi/abs/10.1126/science.ade2574\nhttps://www.biorxiv.org/content/10.1101/2022.07.20.500902v2" }, { "name": "OmegaPLM", "organization": "Massachusetts Institute of Technology (MIT),Westlake University", "orgCategory": "Academia,Academia", "country": "United States, China", "countries": [ "United States", "China" ], "date": "2022-07-22", "year": 2022, "decimalYear": 2022.553425, "domains": [ "Biology" ], "task": "Proteins,Protein folding prediction", "parameters": 670000000, "computeFlop": 1.03514112e+22, "log10Compute": 22.015, "trainingCostUsd2023": 52400.32118528479, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://www.biorxiv.org/content/10.1101/2022.07.21.500999v1" }, { "name": "AlexaTM 20B", "organization": "Amazon", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-08-02", "year": 2022, "decimalYear": 2022.583562, "domains": [ "Language" ], "task": "Language modeling,Translation,Question answering", "parameters": 19750000000, "computeFlop": 2.04374016e+23, "log10Compute": 23.31, "trainingCostUsd2023": 267943.21130997164, "powerDrawW": 102628.10792703854, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Confident", "link": "https://arxiv.org/abs/2208.01448" }, { "name": "GLM-130B", "organization": "Tsinghua University", "orgCategory": "Academia", "country": "China", "countries": [ "China" ], "date": "2022-08-04", "year": 2022, "decimalYear": 2022.589041, "domains": [ "Language" ], "task": "Language modeling/generation,Translation", "parameters": 130000000000, "computeFlop": 3.5490054945e+23, "log10Compute": 23.55, "trainingCostUsd2023": 820296.6313095269, "powerDrawW": 615741.2226117075, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2210.02414" }, { "name": "BlenderBot 3", "organization": "McGill University,Meta AI,Mila - Quebec AI (originally Montreal Institute for Learning Algorithms)", "orgCategory": "Academia,Industry,Academia", "country": "Canada, United States", "countries": [ "Canada", "United States" ], "date": "2022-08-10", "year": 2022, "decimalYear": 2022.605479, "domains": [ "Language" ], "task": "Chat", "parameters": 175000000000, "computeFlop": 4.3e+23, "log10Compute": 23.633, "trainingCostUsd2023": null, "powerDrawW": 102609.82584809877, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Likely", "link": "https://arxiv.org/abs/2208.03188, https://github.com/facebookresearch/ParlAI/blob/main/parlai/zoo/bb3/model_card.md\n\ntraining code: https://parl.ai/projects/bb3/" }, { "name": "BEIT-3", "organization": "Microsoft", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-08-22", "year": 2022, "decimalYear": 2022.638356, "domains": [ "Multimodal", "Vision", "Language" ], "task": "Object detection,Semantic segmentation,Image classification,Visual question answering,Image captioning,Language generation", "parameters": 1900000000, "computeFlop": 70000000000000000000, "log10Compute": 19.845, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2208.10442" }, { "name": "PaLI", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-09-14", "year": 2022, "decimalYear": 2022.70137, "domains": [ "Language", "Vision", "Multimodal" ], "task": "Visual question answering,Language modeling/generation,Image captioning", "parameters": 16900000000, "computeFlop": 1.69e+23, "log10Compute": 23.228, "trainingCostUsd2023": 50878.10777366616, "powerDrawW": 697203.184254025, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2209.06794v4" }, { "name": "Whisper", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-09-21", "year": 2022, "decimalYear": 2022.720548, "domains": [ "Speech" ], "task": "Speech recognition (ASR)", "parameters": 1550000000, "computeFlop": 4.2072663e+21, "log10Compute": 21.624, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://cdn.openai.com/papers/whisper.pdf\n\nhttps://arxiv.org/abs/2212.04356" }, { "name": "DiffDock", "organization": "Massachusetts Institute of Technology (MIT)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2022-10-04", "year": 2022, "decimalYear": 2022.756164, "domains": [ "Biology" ], "task": "Proteins", "parameters": 20240000, "computeFlop": 72000000000000000000, "log10Compute": 19.857, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2210.01776, https://docs.nvidia.com/bionemo-framework/latest/models/diffdock.html" }, { "name": "AlphaTensor", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2022-10-05", "year": 2022, "decimalYear": 2022.758904, "domains": [ "Other", "Games", "Mathematics" ], "task": null, "parameters": null, "computeFlop": 714147840000000000000, "log10Compute": 20.854, "trainingCostUsd2023": null, "powerDrawW": 57646.09262932102, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://www.nature.com/articles/s41586-022-05172-4" }, { "name": "GenSLM", "organization": "University of Chicago,NVIDIA,Harvard University,Cerebras Systems,Technical University of Munich,California Institute of Technology", "orgCategory": "Academia,Industry,Academia,Industry,Academia,Academia", "country": "United States, Germany", "countries": [ "United States", "Germany" ], "date": "2022-10-11", "year": 2022, "decimalYear": 2022.775342, "domains": [ "Biology" ], "task": "Protein or nucleotide language model (pLM/nLM)", "parameters": 25000000000, "computeFlop": 1.42e+21, "log10Compute": 21.152, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://www.biorxiv.org/content/biorxiv/early/2022/10/11/2022.10.10.511571.full.pdf" }, { "name": "U-PaLM (540B)", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-10-20", "year": 2022, "decimalYear": 2022.8, "domains": [ "Language" ], "task": "Language generation,Language modeling/generation,Question answering,Mathematical reasoning,Quantitative reasoning", "parameters": 540000000000, "computeFlop": 2.53e+24, "log10Compute": 24.403, "trainingCostUsd2023": null, "powerDrawW": 348322.23114529514, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2210.11399" }, { "name": "Flan-PaLM 540B", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-10-20", "year": 2022, "decimalYear": 2022.8, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Mathematical reasoning", "parameters": 540000000000, "computeFlop": 2.540000000001e+24, "log10Compute": 24.405, "trainingCostUsd2023": null, "powerDrawW": 348322.23114529514, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2210.11416" }, { "name": "eDiff-I", "organization": "NVIDIA", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-11-02", "year": 2022, "decimalYear": 2022.835616, "domains": [ "Image generation" ], "task": "Image generation,Text-to-image", "parameters": 9100000000, "computeFlop": 54600000000000000000, "log10Compute": 19.737, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2211.01324" }, { "name": "Mogrifier RLSTM (WT2)", "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2022-11-03", "year": 2022, "decimalYear": 2022.838356, "domains": [ "Language" ], "task": "Language modeling", "parameters": 35000000, "computeFlop": 140000000000000000, "log10Compute": 17.146, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2211.01848" }, { "name": "InternImage", "organization": "Shanghai AI Lab,Tsinghua University,Nanjing University,SenseTime,Chinese University of Hong Kong (CUHK)", "orgCategory": "Academia,Academia,Academia,Industry,Academia", "country": "China, Hong Kong", "countries": [ "China", "Hong Kong" ], "date": "2022-11-10", "year": 2022, "decimalYear": 2022.857534, "domains": [ "Vision" ], "task": "Image classification,Object detection,Image segmentation", "parameters": 1080000000, "computeFlop": 2.408e+21, "log10Compute": 21.382, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2211.05778" }, { "name": "EVA-01", "organization": "Beijing Academy of Artificial Intelligence / BAAI,Huazhong University of Science and Technology,Zhejiang University (ZJU),Beijing Institute of Technology", "orgCategory": "Academia,Academia,Academia,Academia", "country": "China", "countries": [ "China" ], "date": "2022-11-14", "year": 2022, "decimalYear": 2022.868493, "domains": [ "Vision" ], "task": "Image classification,Object detection,Semantic segmentation,Video classification", "parameters": 1011000000, "computeFlop": 1.501e+22, "log10Compute": 22.176, "trainingCostUsd2023": 29374.46909439904, "powerDrawW": 102390.69476171424, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2211.07636" }, { "name": "Galactica", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2022-11-16", "year": 2022, "decimalYear": 2022.873973, "domains": [ "Language", "Biology" ], "task": "Language modeling,Question answering,Mathematical reasoning,Medical diagnosis,Language modeling/generation", "parameters": 120000000000, "computeFlop": 3.24e+23, "log10Compute": 23.511, "trainingCostUsd2023": 591076.8943544837, "powerDrawW": 102386.13451047534, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Likely", "link": "https://arxiv.org/abs/2211.09085" }, { "name": "Fusion in Encoder", "organization": "Samsung", "orgCategory": "Industry", "country": "South Korea", "countries": [ "South Korea" ], "date": "2022-11-18", "year": 2022, "decimalYear": 2022.879452, "domains": [ "Language" ], "task": "Question answering,Language modeling/generation", "parameters": 330000000, "computeFlop": 130000000000000000000, "log10Compute": 20.114, "trainingCostUsd2023": 233.0630322095263, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2211.10147" }, { "name": "AR-LDM", "organization": "Alibaba,University of Waterloo,Vector Institute", "orgCategory": "Industry,Academia,Academia", "country": "China, Canada", "countries": [ "China", "Canada" ], "date": "2022-11-20", "year": 2022, "decimalYear": 2022.884932, "domains": [ "Image generation" ], "task": "Text-to-image", "parameters": 1500000000, "computeFlop": 510000000000000000000, "log10Compute": 20.708, "trainingCostUsd2023": 745.8360575556148, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2211.10950" }, { "name": "Discriminator Guidance", "organization": "Korea Advanced Institute of Science and Technology (KAIST),NAVER", "orgCategory": "Academia,Industry", "country": "South Korea", "countries": [ "South Korea" ], "date": "2022-11-28", "year": 2022, "decimalYear": 2022.906849, "domains": [ "Image generation" ], "task": "Image generation", "parameters": null, "computeFlop": 215700000010000000000, "log10Compute": 20.334, "trainingCostUsd2023": 337.8811363173893, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2211.17091v4" }, { "name": "Vega v2", "organization": "Wuhan University,JD Explore Academy,Shanghai AI Lab,Nanyang Technological University,Washington University in St Louis,Chongqing University of Posts and Telecommunications,University of Sydney", "orgCategory": "Academia,Academia,Academia,Academia,Academia,Academia", "country": "China, Singapore, United States, Australia", "countries": [ "China", "Singapore", "United States", "Australia" ], "date": "2022-12-04", "year": 2022, "decimalYear": 2022.923288, "domains": [ "Language" ], "task": "Language modeling,Question answering,Word sense disambiguation", "parameters": 6000000000, "computeFlop": 7.76e+22, "log10Compute": 22.89, "trainingCostUsd2023": null, "powerDrawW": 255862.7534697245, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2212.01853" }, { "name": "CaLM", "organization": "University of Oxford", "orgCategory": "Academia", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2022-12-19", "year": 2022, "decimalYear": 2022.964384, "domains": [ "Biology" ], "task": "Protein or nucleotide language model (pLM/nLM),Protein embedding,Protein property prediction,Protein localization prediction", "parameters": 86000000, "computeFlop": 29000000000000000000, "log10Compute": 19.462, "trainingCostUsd2023": null, "powerDrawW": 1278.886496016633, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://www.biorxiv.org/content/10.1101/2022.12.15.519894v1.full.pdf" }, { "name": "Hybrid H3-2.7B", "organization": "Stanford University,University at Buffalo", "orgCategory": "Academia,Academia", "country": "United States", "countries": [ "United States" ], "date": "2022-12-28", "year": 2022, "decimalYear": 2022.989041, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering", "parameters": 2700000000, "computeFlop": 6.48e+21, "log10Compute": 21.812, "trainingCostUsd2023": null, "powerDrawW": 6393.151008587547, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2212.14052" }, { "name": "VALL-E", "organization": "Microsoft", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-01-05", "year": 2023, "decimalYear": 2023.010959, "domains": [ "Audio", "Speech" ], "task": "Speech synthesis,Text-to-speech (TTS)", "parameters": 353000000, "computeFlop": 10100000000000000000, "log10Compute": 19.004, "trainingCostUsd2023": 11.40575196486363, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://arxiv.org/abs/2301.02111" }, { "name": "DreamerV3", "organization": "DeepMind,University of Toronto", "orgCategory": "Industry,Academia", "country": "United Kingdom, Canada", "countries": [ "United Kingdom", "Canada" ], "date": "2023-01-10", "year": 2023, "decimalYear": 2023.024658, "domains": [ "Games" ], "task": "Open ended play", "parameters": 200000000, "computeFlop": 220320000000000000000, "log10Compute": 20.343, "trainingCostUsd2023": null, "powerDrawW": 9586.950671364422, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2301.04104v1" }, { "name": "Nucleotide Transformer", "organization": "NVIDIA,Technical University of Munich,InstaDeep", "orgCategory": "Industry,Academia,Industry", "country": "United States, Germany, United Kingdom", "countries": [ "United States", "Germany", "United Kingdom" ], "date": "2023-01-15", "year": 2023, "decimalYear": 2023.038356, "domains": [ "Biology" ], "task": "Protein or nucleotide language model (pLM/nLM),Nucleotide generation", "parameters": 2500000000, "computeFlop": 8.08e+21, "log10Compute": 21.907, "trainingCostUsd2023": 51064.70621394041, "powerDrawW": 102249.42137571001, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Likely", "link": "https://www.biorxiv.org/content/10.1101/2023.01.11.523679v1.full.pdf" }, { "name": "Ankh_large", "organization": "Technical University of Munich,Columbia University", "orgCategory": "Academia,Academia", "country": "Germany, United States", "countries": [ "Germany", "United States" ], "date": "2023-01-16", "year": 2023, "decimalYear": 2023.041096, "domains": [ "Biology" ], "task": "Protein generation,Proteins,Protein or nucleotide language model (pLM/nLM),Protein contact and distance prediction,Protein classification,Protein localization prediction,Protein fold classification", "parameters": 1900000000, "computeFlop": 6.5e+21, "log10Compute": 21.813, "trainingCostUsd2023": 4802.398249072418, "powerDrawW": 43455.036357563025, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2301.06568" }, { "name": "DDPM-IP (CelebA)", "organization": "Utrecht University", "orgCategory": "Academia", "country": "Netherlands", "countries": [ "Netherlands" ], "date": "2023-01-27", "year": 2023, "decimalYear": 2023.071233, "domains": [ "Image generation" ], "task": "Image generation,Text-to-image", "parameters": 295000000, "computeFlop": 350000000000000000000, "log10Compute": 20.544, "trainingCostUsd2023": 390.4861317667304, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2301.11706v3" }, { "name": "BLIP-2 (Q-Former)", "organization": "Salesforce Research", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-01-30", "year": 2023, "decimalYear": 2023.079452, "domains": [ "Vision", "Language" ], "task": "Visual question answering,Image captioning", "parameters": 1480000000, "computeFlop": 1.20000000001e+21, "log10Compute": 21.079, "trainingCostUsd2023": 1960.8225382725811, "powerDrawW": 12776.908953102382, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2301.12597" }, { "name": "ViT-22B", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-02-10", "year": 2023, "decimalYear": 2023.109589, "domains": [ "Vision" ], "task": "Object detection,Image classification", "parameters": 21743000000, "computeFlop": 1.93248e+23, "log10Compute": 23.286, "trainingCostUsd2023": 285555.57016183576, "powerDrawW": 694893.6029178143, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2302.05442v1" }, { "name": "LLaMA-65B", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-02-24", "year": 2023, "decimalYear": 2023.147945, "domains": [ "Language" ], "task": "Language modeling,Code generation", "parameters": 65200000000, "computeFlop": 5.5e+23, "log10Compute": 23.74, "trainingCostUsd2023": 578026.3043, "powerDrawW": 1634534.091472035, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2302.13971" }, { "name": "DiT-XL/2", "organization": "New York University (NYU),University of California (UC) Berkeley", "orgCategory": "Academia,Academia", "country": "United States", "countries": [ "United States" ], "date": "2023-03-02", "year": 2023, "decimalYear": 2023.164384, "domains": [ "Image generation" ], "task": "Image generation", "parameters": 675000000, "computeFlop": 600000000000000000000, "log10Compute": 20.778, "trainingCostUsd2023": 111048.19613085664, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2212.09748" }, { "name": "AudioGen", "organization": "Meta AI,Hebrew University of Jerusalem", "orgCategory": "Industry,Academia", "country": "United States, Israel", "countries": [ "United States", "Israel" ], "date": "2023-03-05", "year": 2023, "decimalYear": 2023.172603, "domains": [ "Audio" ], "task": "Audio generation", "parameters": 1000000000, "computeFlop": 9.5e+21, "log10Compute": 21.978, "trainingCostUsd2023": 9429.74091062958, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Likely", "link": "https://arxiv.org/abs/2209.15352" }, { "name": "GPT-4 (Mar 2023)", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-03-15", "year": 2023, "decimalYear": 2023.2, "domains": [ "Multimodal", "Language", "Vision" ], "task": "Language modeling,Language modeling/generation,Question answering,Visual question answering", "parameters": 1800000000000, "computeFlop": 2.1e+25, "log10Compute": 25.322, "trainingCostUsd2023": 37334304.976625234, "powerDrawW": 19944368.12599428, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Likely", "link": "https://arxiv.org/abs/2303.08774" }, { "name": "Falcon-40B", "organization": "Technology Innovation Institute", "orgCategory": "Government", "country": "UAE", "countries": [ "UAE" ], "date": "2023-03-15", "year": 2023, "decimalYear": 2023.2, "domains": [ "Language" ], "task": "Language modeling,Language modeling/generation,Question answering", "parameters": 40000000000, "computeFlop": 2.4e+23, "log10Compute": 23.38, "trainingCostUsd2023": 319783.157242365, "powerDrawW": 306345.49441527214, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2311.16867; https://www.tii.ae/news/abu-dhabi-based-technology-innovation-institute-introduces-falcon-llm-foundational-large" }, { "name": "PanGu-Σ", "organization": "Huawei Noah's Ark Lab", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2023-03-20", "year": 2023, "decimalYear": 2023.213699, "domains": [ "Language" ], "task": "Code generation,Language modeling,Translation,Question answering", "parameters": 1085000000000, "computeFlop": 4.67e+23, "log10Compute": 23.669, "trainingCostUsd2023": null, "powerDrawW": 316521.76523003745, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2303.10845" }, { "name": "SigLIP 400M", "organization": "Google DeepMind", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-03-27", "year": 2023, "decimalYear": 2023.232877, "domains": [ "Vision" ], "task": "Image classification,Image embedding", "parameters": 400000000, "computeFlop": 4.9467301e+21, "log10Compute": 21.694, "trainingCostUsd2023": null, "powerDrawW": 21693.674492506372, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2303.15343" }, { "name": "VideoMAE V2", "organization": "Nanjing University,Shenzhen Institute of Advanced Technology,Shanghai AI Lab", "orgCategory": "Academia,Academia", "country": "China", "countries": [ "China" ], "date": "2023-03-29", "year": 2023, "decimalYear": 2023.238356, "domains": [ "Video" ], "task": "Action recognition", "parameters": 1000000000, "computeFlop": 9.7e+21, "log10Compute": 21.987, "trainingCostUsd2023": 18339.96928068276, "powerDrawW": 51041.66660009344, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2303.16727v2" }, { "name": "BloombergGPT", "organization": "Bloomberg,Johns Hopkins University", "orgCategory": "Industry,Academia", "country": "United States", "countries": [ "United States" ], "date": "2023-03-30", "year": 2023, "decimalYear": 2023.241096, "domains": [ "Language" ], "task": "Language modeling,Language modeling/generation,Question answering,Financial management,Text classification", "parameters": 50558868480, "computeFlop": 2.36e+23, "log10Compute": 23.373, "trainingCostUsd2023": 369586.1352802876, "powerDrawW": 408324.2395754059, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2303.17564" }, { "name": "Segment Anything Model", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-04-05", "year": 2023, "decimalYear": 2023.257534, "domains": [ "Vision" ], "task": "Image segmentation", "parameters": 636000000, "computeFlop": 7.8e+21, "log10Compute": 21.892, "trainingCostUsd2023": 15888.411228475235, "powerDrawW": 204134.84223782964, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2304.02643" }, { "name": "Incoder-6.7B", "organization": "Facebook AI Research,University of Washington,University of California (UC) Berkeley,Carnegie Mellon University (CMU),Toyota Technological Institute at Chicago", "orgCategory": "Industry,Academia,Academia,Academia,Academia", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2023-04-09", "year": 2023, "decimalYear": 2023.268493, "domains": [ "Language" ], "task": "Code generation", "parameters": 6700000000, "computeFlop": 3.00001e+21, "log10Compute": 21.477, "trainingCostUsd2023": 3129.0771365574788, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2204.05999" }, { "name": "DINOv2", "organization": "Facebook AI Research,INRIA", "orgCategory": "Industry,Academia", "country": "United States, France", "countries": [ "United States", "France" ], "date": "2023-04-14", "year": 2023, "decimalYear": 2023.282192, "domains": [ "Vision" ], "task": "Image representation,Image classification", "parameters": 1140000000, "computeFlop": 7.41851136e+21, "log10Compute": 21.87, "trainingCostUsd2023": 10203.605181058361, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2304.07193" }, { "name": "LLaVA", "organization": "University of Wisconsin Madison,Microsoft Research,Columbia University", "orgCategory": "Academia,Industry,Academia", "country": "United States", "countries": [ "United States" ], "date": "2023-04-17", "year": 2023, "decimalYear": 2023.290411, "domains": [ "Multimodal", "Vision", "Language" ], "task": "Chat,Question answering,Visual question answering", "parameters": 13000000000, "computeFlop": 7.8049e+22, "log10Compute": 22.892, "trainingCostUsd2023": 42.46267260692187, "powerDrawW": 6377.509314719864, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2304.08485" }, { "name": "StarCoder", "organization": "Hugging Face,ServiceNow,Northeastern University,Mila - Quebec AI (originally Montreal Institute for Learning Algorithms),Carnegie Mellon University (CMU),Johns Hopkins University,Leipzig University,ScaDS.AI,Queen Mary University of London,Roblox,Sea AI Lab,Technion - Israel Institute of Technology,Monash University,CSIRO,Data61,McGill University,Saama,University of British Columbia (UBC),Massachusetts Institute of Technology (MIT),Technical University of Munich,IBM,University of Vermont,UnfoldML,SAP,University of Notre Dame,Columbia University,New York University (NYU),University of Allahabad,Discover Dollar,Toloka,Telefonica,Stanford University,Weizmann Institute of Science,Alan Turing Institute,Wellesley College,EleutherAI,Forschungszentrum Julich", "orgCategory": "Industry,Industry,Academia,Academia,Academia,Academia,Academia,Academia,Industry,Academia,Academia,Government,Government,Academia,Academia,Academia,Academia,Industry,Academia,Industry,Academia,Academia,Academia,Academia,Industry,Industry,Industry,Academia,Academia,Government,Academia,Research collective,Government", "country": "United States, Canada, Germany, United Kingdom, Singapore, Israel, Australia, Sweden, India, Netherlands, Spain", "countries": [ "United States", "Canada", "Germany", "United Kingdom", "Singapore", "Israel", "Australia", "Sweden", "India", "Netherlands", "Spain" ], "date": "2023-05-09", "year": 2023, "decimalYear": 2023.350685, "domains": [ "Language" ], "task": "Code generation", "parameters": 15500000000, "computeFlop": 8.46e+22, "log10Compute": 22.927, "trainingCostUsd2023": 212217.65075330864, "powerDrawW": 407960.6765621668, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://arxiv.org/abs/2305.06161" }, { "name": "PaLM 2", "organization": "Google", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-05-10", "year": 2023, "decimalYear": 2023.353425, "domains": [ "Language" ], "task": "Language modeling,Language modeling/generation", "parameters": 340000000000, "computeFlop": 7.34e+24, "log10Compute": 24.866, "trainingCostUsd2023": 5014266.983518871, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Likely", "link": "https://arxiv.org/abs/2305.10403" }, { "name": "InstructBLIP", "organization": "Salesforce Research,Hong Kong University of Science and Technology (HKUST),Nanyang Technological University", "orgCategory": "Industry,Academia,Academia", "country": "United States, Hong Kong, Singapore", "countries": [ "United States", "Hong Kong", "Singapore" ], "date": "2023-05-11", "year": 2023, "decimalYear": 2023.356164, "domains": [ "Multimodal", "Language", "Vision" ], "task": "Visual question answering,Chat", "parameters": 13000000000, "computeFlop": 194000000000000000000, "log10Compute": 20.288, "trainingCostUsd2023": null, "powerDrawW": 12748.20334097663, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2305.06500" }, { "name": "ONE-PEACE", "organization": "Alibaba,Huazhong University of Science and Technology", "orgCategory": "Industry,Academia", "country": "China", "countries": [ "China" ], "date": "2023-05-18", "year": 2023, "decimalYear": 2023.375342, "domains": [ "Multimodal", "Vision", "Speech", "Language" ], "task": "Image classification,Speech recognition (ASR),Audio question answering,Audio classification,Semantic segmentation", "parameters": 4000000000, "computeFlop": 180000000000000000000, "log10Compute": 20.255, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Speculative", "link": "https://arxiv.org/abs/2305.11172v1" }, { "name": "PaLI-X", "organization": "Google Research", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-05-29", "year": 2023, "decimalYear": 2023.405479, "domains": [ "Multimodal", "Language", "Vision", "Video" ], "task": "Image captioning,Video description,Character recognition (OCR),Visual question answering", "parameters": 55000000000, "computeFlop": 5.6e+23, "log10Compute": 23.748, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2305.18565" }, { "name": "GPT-4 (Jun 2023)", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-06-13", "year": 2023, "decimalYear": 2023.446575, "domains": [ "Multimodal", "Language", "Vision" ], "task": "Language modeling,Language modeling/generation,Question answering,Visual question answering", "parameters": 1800000000000, "computeFlop": 2.1e+25, "log10Compute": 25.322, "trainingCostUsd2023": null, "powerDrawW": 19904434.788947295, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Likely", "link": "https://arxiv.org/abs/2303.08774" }, { "name": "HyenaDNA", "organization": "Stanford University,Harvard University,Mila - Quebec AI (originally Montreal Institute for Learning Algorithms),University of Montreal / Université de Montréal", "orgCategory": "Academia,Academia,Academia,Academia", "country": "United States, Canada", "countries": [ "United States", "Canada" ], "date": "2023-06-27", "year": 2023, "decimalYear": 2023.484932, "domains": [ "Biology" ], "task": "Protein or nucleotide language model (pLM/nLM)", "parameters": 6600000, "computeFlop": 1.811e+21, "log10Compute": 21.258, "trainingCostUsd2023": 5000, "powerDrawW": 6367.433640556241, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://arxiv.org/abs/2306.15794" }, { "name": "Pangu-Weather", "organization": "Huawei", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2023-07-05", "year": 2023, "decimalYear": 2023.506849, "domains": [ "Earth science" ], "task": "Weather forecasting", "parameters": 256000000, "computeFlop": 3.98e+22, "log10Compute": 22.6, "trainingCostUsd2023": 51279.017519054316, "powerDrawW": 114593.38832967015, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://www.nature.com/articles/s41586-023-06185-3, https://www.huaweicloud.com/intl/en-us/news/20230707180809498.html,\nhttps://www.huawei.com/en/news/2023/7/pangu-ai-model-nature-publish" }, { "name": "xTrimoPGLM -100B", "organization": "Tsinghua University,BioMap Research", "orgCategory": "Academia,Industry", "country": "China", "countries": [ "China" ], "date": "2023-07-06", "year": 2023, "decimalYear": 2023.509589, "domains": [ "Biology" ], "task": "Proteins,Protein or nucleotide language model (pLM/nLM),Protein generation", "parameters": 100000000000, "computeFlop": 6.2e+23, "log10Compute": 23.792, "trainingCostUsd2023": 1823415.258, "powerDrawW": 611151.12765533, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://www.biorxiv.org/content/10.1101/2023.07.05.547496v4" }, { "name": "InternLM", "organization": "Shanghai AI Lab,SenseTime", "orgCategory": "Academia,Industry", "country": "China, Hong Kong", "countries": [ "China", "Hong Kong" ], "date": "2023-07-06", "year": 2023, "decimalYear": 2023.509589, "domains": [ "Language" ], "task": "Language modeling", "parameters": 104000000000, "computeFlop": 9.984e+23, "log10Compute": 23.999, "trainingCostUsd2023": 1505257.378, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://github.com/InternLM/InternLM-techreport/blob/main/InternLM.pdf" }, { "name": "Claude 2", "organization": "Anthropic", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-07-11", "year": 2023, "decimalYear": 2023.523288, "domains": [ "Language" ], "task": "Language modeling,Chat,Language modeling/generation,Question answering", "parameters": null, "computeFlop": 3.866e+24, "log10Compute": 24.587, "trainingCostUsd2023": 4902644.123383027, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Speculative", "link": "https://www.anthropic.com/index/claude-2, https://www-files.anthropic.com/production/images/Model-Card-Claude-2.pdf" }, { "name": "Llama 2-70B", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-07-18", "year": 2023, "decimalYear": 2023.542466, "domains": [ "Language" ], "task": "Language modeling,Language modeling/generation,Question answering", "parameters": 70000000000, "computeFlop": 8.1e+23, "log10Compute": 23.908, "trainingCostUsd2023": 1102561.194, "powerDrawW": 795557.0703894598, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://ai.meta.com/research/publications/llama-2-open-foundation-and-fine-tuned-chat-models/\nhttps://arxiv.org/abs/2307.09288" }, { "name": "Llama 2-7B", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-07-18", "year": 2023, "decimalYear": 2023.542466, "domains": [ "Language" ], "task": "Language modeling", "parameters": 7000000000, "computeFlop": 8.4e+22, "log10Compute": 22.924, "trainingCostUsd2023": 114259.38527188865, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://ai.meta.com/research/publications/llama-2-open-foundation-and-fine-tuned-chat-models/" }, { "name": "AudioLM", "organization": "Google Research", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-07-26", "year": 2023, "decimalYear": 2023.564384, "domains": [ "Audio" ], "task": "Audio generation", "parameters": 1500000000, "computeFlop": 3900000000000000000, "log10Compute": 18.591, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://arxiv.org/abs/2209.03143" }, { "name": "GGNN", "organization": "Westlake University,Tsinghua University,Toyota Technological Institute at Chicago", "orgCategory": "Academia,Academia,Academia", "country": "China, United States", "countries": [ "China", "United States" ], "date": "2023-08-05", "year": 2023, "decimalYear": 2023.591781, "domains": [ "Biology" ], "task": "Proteins,Protein interaction prediction,Protein protein binding affinity prediction,Protein representation learning", "parameters": null, "computeFlop": 7.56e+21, "log10Compute": 21.879, "trainingCostUsd2023": null, "powerDrawW": 1590.4764725839211, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://www.nature.com/articles/s42003-023-05133-1" }, { "name": "PeptideBERT", "organization": "Carnegie Mellon University (CMU)", "orgCategory": "Academia", "country": "United States", "countries": [ "United States" ], "date": "2023-08-28", "year": 2023, "decimalYear": 2023.654795, "domains": [ "Biology" ], "task": "Proteins,Protein property prediction", "parameters": null, "computeFlop": 49000000000000000, "log10Compute": 16.69, "trainingCostUsd2023": null, "powerDrawW": 272.95021398005304, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2309.03099" }, { "name": "Jais", "organization": "Cerebras Systems,Mohamed bin Zayed University of Artificial Intelligence (MBZUAI),Inception G42", "orgCategory": "Industry,Academia,Industry", "country": "United States, UAE", "countries": [ "United States", "UAE" ], "date": "2023-08-29", "year": 2023, "decimalYear": 2023.657534, "domains": [ "Language" ], "task": "Language modeling,Chat,Translation,Language modeling/generation,Question answering", "parameters": 13000000000, "computeFlop": 4.8946763e+22, "log10Compute": 22.69, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2308.16149" }, { "name": "Swift", "organization": "Intel Labs", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-08-30", "year": 2023, "decimalYear": 2023.660274, "domains": [ "Robotics" ], "task": "Helicopter driving", "parameters": 56804, "computeFlop": 53370000000000000, "log10Compute": 16.727, "trainingCostUsd2023": null, "powerDrawW": 382.113280348051, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://www.nature.com/articles/s41586-023-06419-4" }, { "name": "Falcon-180B", "organization": "Technology Innovation Institute", "orgCategory": "Government", "country": "UAE", "countries": [ "UAE" ], "date": "2023-09-06", "year": 2023, "decimalYear": 2023.679452, "domains": [ "Language" ], "task": "Language modeling,Language modeling/generation,Question answering", "parameters": 180000000000, "computeFlop": 3.76e+24, "log10Compute": 24.575, "trainingCostUsd2023": 10743500.868805483, "powerDrawW": 3254975.4289709153, "frontier": true, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://falconllm.tii.ae/falcon-180b.html; https://arxiv.org/abs/2311.16867" }, { "name": "Amazon Titan", "organization": "Amazon", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-09-28", "year": 2023, "decimalYear": 2023.739726, "domains": [ "Language", "Image generation" ], "task": "Semantic search,Image generation,Language modeling/generation,Code generation,Chat,Text-to-image,Translation", "parameters": 200000000000, "computeFlop": 4.8e+24, "log10Compute": 24.681, "trainingCostUsd2023": 7933464.673729055, "powerDrawW": 10929327.206416877, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Likely", "link": "https://aws.amazon.com/bedrock/titan/" }, { "name": "FinGPT-13B", "organization": "University of California Los Angeles (UCLA),Columbia University,New York University (NYU)", "orgCategory": "Academia,Academia,Academia", "country": "United States", "countries": [ "United States" ], "date": "2023-10-07", "year": 2023, "decimalYear": 2023.764384, "domains": [ "Language" ], "task": "Named entity recognition (NER),Sentiment classification,Language modeling/generation,Financial management", "parameters": 13000000000, "computeFlop": 1.6e+23, "log10Compute": 23.204, "trainingCostUsd2023": null, "powerDrawW": 381.79005907274245, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2310.04793; https://github.com/AI4Finance-Foundation/FinGPT" }, { "name": "RoseTTAFold All-Atom (RFAA)", "organization": "University of Washington,Seoul National University,University of Sheffield", "orgCategory": "Academia,Academia,Academia", "country": "United States, South Korea, United Kingdom", "countries": [ "United States", "South Korea", "United Kingdom" ], "date": "2023-10-09", "year": 2023, "decimalYear": 2023.769863, "domains": [ "Biology" ], "task": "Protein folding prediction,Proteins", "parameters": null, "computeFlop": 214000000000000000000, "log10Compute": 20.33, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://www.biorxiv.org/content/10.1101/2023.10.09.561603v1" }, { "name": "CODEFUSION (Python)", "organization": "Microsoft,Microsoft Research", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-10-26", "year": 2023, "decimalYear": 2023.816438, "domains": [ "Language" ], "task": "Code generation", "parameters": 75000000, "computeFlop": 7920000000000000000, "log10Compute": 18.899, "trainingCostUsd2023": 8.542235671062665, "powerDrawW": 2381.36215809478, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2310.17680 (was withdrawn)\n\nhttps://www.microsoft.com/en-us/research/wp-content/uploads/2023/11/CodeFusion-Revised-CameraReady.pdf" }, { "name": "ChatGLM3-6B", "organization": "Z.ai (Zhipu AI)", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2023-10-27", "year": 2023, "decimalYear": 2023.819178, "domains": [ "Multimodal", "Language", "Vision" ], "task": "Chat,Visual question answering,Code generation", "parameters": 6000000000, "computeFlop": 5.04e+22, "log10Compute": 22.702, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Speculative", "link": "https://www.zhipuai.cn/en/news/76\n\nhttps://huggingface.co/zai-org/chatglm3-6b" }, { "name": "Skywork-13B", "organization": "Kunlun Inc.", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2023-10-30", "year": 2023, "decimalYear": 2023.827397, "domains": [ "Language" ], "task": "Language modeling,Language modeling/generation,Translation", "parameters": 13000000000, "computeFlop": 2.5e+23, "log10Compute": 23.398, "trainingCostUsd2023": null, "powerDrawW": 253989.3377909499, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://arxiv.org/abs/2310.19341" }, { "name": "Yi-34B", "organization": "01.AI", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2023-11-02", "year": 2023, "decimalYear": 2023.835616, "domains": [ "Language" ], "task": "Chat,Language modeling/generation,Translation,Code generation", "parameters": 34000000000, "computeFlop": 6.1e+23, "log10Compute": 23.785, "trainingCostUsd2023": null, "powerDrawW": 101588.94792366635, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://arxiv.org/abs/2403.04652" }, { "name": "Grok-1", "organization": "xAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-11-04", "year": 2023, "decimalYear": 2023.841096, "domains": [ "Language" ], "task": "Language modeling,Chat", "parameters": 314000000000, "computeFlop": 2.90000000001e+24, "log10Compute": 24.462, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://x.ai/model-card/, https://x.ai/blog/grok-os" }, { "name": "LLaVA 1.5", "organization": "University of Wisconsin Madison,Microsoft Research", "orgCategory": "Academia,Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-11-05", "year": 2023, "decimalYear": 2023.843836, "domains": [ "Multimodal", "Language", "Vision" ], "task": "Chat,Question answering,Visual question answering", "parameters": 13000000000, "computeFlop": 7.807e+22, "log10Compute": 22.892, "trainingCostUsd2023": null, "powerDrawW": 6348.88507402357, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://arxiv.org/abs/2310.03744," }, { "name": "CogVLM-17B", "organization": "Tsinghua University,Z.ai (Zhipu AI),Beihang University", "orgCategory": "Academia,Industry,Academia", "country": "China", "countries": [ "China" ], "date": "2023-11-06", "year": 2023, "decimalYear": 2023.846575, "domains": [ "Multimodal", "Vision", "Language" ], "task": "Image captioning,Visual question answering,Chat", "parameters": 17000000000, "computeFlop": 6.331e+22, "log10Compute": 22.801, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://arxiv.org/abs/2311.03079\nhttps://huggingface.co/THUDM/cogvlm-chat-hf\nhttps://github.com/THUDM/CogVLM" }, { "name": "RoFormer", "organization": "Zhuiyi Technology", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2023-11-08", "year": 2023, "decimalYear": 2023.852055, "domains": [ "Language" ], "task": "Text classification,Language modeling/generation", "parameters": 110000000, "computeFlop": 2162688000000000000, "log10Compute": 18.335, "trainingCostUsd2023": null, "powerDrawW": 2380.6728491831727, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2104.09864" }, { "name": "MultiBand Diffusion", "organization": "Meta AI,Hebrew University of Jerusalem,LORIA", "orgCategory": "Industry,Academia,Academia", "country": "United States, Israel, France", "countries": [ "United States", "Israel", "France" ], "date": "2023-11-08", "year": 2023, "decimalYear": 2023.852055, "domains": [ "Audio", "Speech" ], "task": "Audio generation", "parameters": null, "computeFlop": 26000000000000000000, "log10Compute": 19.415, "trainingCostUsd2023": 22.81032329809286, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2308.02560" }, { "name": "Volcano 13B", "organization": "Korea University,Korea Advanced Institute of Science and Technology (KAIST),LG", "orgCategory": "Academia,Academia,Industry", "country": "South Korea", "countries": [ "South Korea" ], "date": "2023-11-13", "year": 2023, "decimalYear": 2023.865753, "domains": [ "Language", "Multimodal", "Vision" ], "task": "Language modeling/generation,Visual question answering", "parameters": 13000000000, "computeFlop": 4.56e+22, "log10Compute": 22.659, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Likely", "link": "https://arxiv.org/abs/2311.07362" }, { "name": "SPHINX (Llama 2 13B)", "organization": "Shanghai AI Lab,Chinese University of Hong Kong (CUHK),ShanghaiTech University", "orgCategory": "Academia,Academia,Academia", "country": "China, Hong Kong", "countries": [ "China", "Hong Kong" ], "date": "2023-11-13", "year": 2023, "decimalYear": 2023.865753, "domains": [ "Vision", "Language", "Multimodal" ], "task": "Visual question answering,Image captioning", "parameters": 19900000000, "computeFlop": 3.04e+22, "log10Compute": 22.483, "trainingCostUsd2023": 239188.6875340231, "powerDrawW": 25391.016357362594, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Likely", "link": "https://arxiv.org/abs/2311.07575" }, { "name": "GraphCast", "organization": "Google DeepMind", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-11-14", "year": 2023, "decimalYear": 2023.868493, "domains": [ "Earth science" ], "task": "Weather forecasting", "parameters": null, "computeFlop": 2.1000000000000002e+22, "log10Compute": 22.322, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Speculative", "link": "https://www.science.org/doi/epdf/10.1126/science.adi2336\n\nhttps://arxiv.org/abs/2212.12794" }, { "name": "Nemotron-3-8B", "organization": "NVIDIA", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-11-15", "year": 2023, "decimalYear": 2023.871233, "domains": [ "Language" ], "task": "Chat,Language generation,Language modeling/generation,Translation,Code generation,Question answering", "parameters": 8000000000, "computeFlop": 1.8e+23, "log10Compute": 23.255, "trainingCostUsd2023": 214467.02013524104, "powerDrawW": 812476.3359553809, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://developer.nvidia.com/blog/nvidia-ai-foundation-models-build-custom-enterprise-chatbots-and-co-pilots-with-production-ready-llms/\n\nhttps://catalog.ngc.nvidia.com/orgs/nvidia/teams/nemo/models/nemotron-3-8b-base-4k" }, { "name": "Inflection-2", "organization": "Inflection AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-11-22", "year": 2023, "decimalYear": 2023.890411, "domains": [ "Language" ], "task": "Language modeling,Language modeling/generation,Chat,Question answering", "parameters": null, "computeFlop": 1.001e+25, "log10Compute": 25, "trainingCostUsd2023": 13461144.182562498, "powerDrawW": 6941464.657305156, "frontier": true, "openWeights": false, "accessibility": "Hosted access (no API)", "confidence": "Confident", "link": "https://inflection.ai/inflection-2" }, { "name": "Qwen-72B", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2023-11-30", "year": 2023, "decimalYear": 2023.912329, "domains": [ "Language" ], "task": "Chat,Code generation,Language modeling/generation,Question answering", "parameters": 72000000000, "computeFlop": 1.3e+24, "log10Compute": 24.114, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://huggingface.co/Qwen/Qwen-72B" }, { "name": "Gemini 1.0 Ultra", "organization": "Google DeepMind", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-12-06", "year": 2023, "decimalYear": 2023.928767, "domains": [ "Multimodal", "Language", "Vision" ], "task": "Language modeling,Visual question answering,Chat,Translation", "parameters": null, "computeFlop": 5.0000000001e+25, "log10Compute": 25.699, "trainingCostUsd2023": 30719419.93534952, "powerDrawW": 38423900.11232233, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Speculative", "link": "https://storage.googleapis.com/deepmind-media/gemini/gemini_1_report.pdf" }, { "name": "Llama Guard", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-12-07", "year": 2023, "decimalYear": 2023.931507, "domains": [ "Language" ], "task": "Chat", "parameters": 7000000000, "computeFlop": 1.6e+23, "log10Compute": 23.204, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://arxiv.org/abs/2312.06674" }, { "name": "Mixtral 8x7B", "organization": "Mistral AI", "orgCategory": "Industry", "country": "France", "countries": [ "France" ], "date": "2023-12-11", "year": 2023, "decimalYear": 2023.942466, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Quantitative reasoning,Translation", "parameters": 46700000000, "computeFlop": 7.74e+23, "log10Compute": 23.889, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Speculative", "link": "https://mistral.ai/news/mixtral-of-experts/, https://arxiv.org/abs/2401.04088" }, { "name": "VILA-13B", "organization": "NVIDIA,Massachusetts Institute of Technology (MIT)", "orgCategory": "Industry,Academia", "country": "United States", "countries": [ "United States" ], "date": "2023-12-12", "year": 2023, "decimalYear": 2023.945205, "domains": [ "Multimodal", "Language", "Vision" ], "task": "Chat,Visual question answering,Image captioning,Language modeling/generation,Question answering", "parameters": 13350839296, "computeFlop": 2.3003136e+21, "log10Compute": 21.362, "trainingCostUsd2023": null, "powerDrawW": 101498.49533614099, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2312.07533\nhttps://huggingface.co/Efficient-Large-Model/VILA-13b" }, { "name": "FunSearch", "organization": "Google DeepMind", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-12-14", "year": 2023, "decimalYear": 2023.950685, "domains": [ "Language", "Search" ], "task": "Code generation", "parameters": 15000000000, "computeFlop": 3.87e+23, "log10Compute": 23.588, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Speculative", "link": "https://www.nature.com/articles/s41586-023-06924-6\nhttps://deepmind.google/discover/blog/funsearch-making-new-discoveries-in-mathematical-sciences-using-large-language-models/" }, { "name": "CogAgent", "organization": "Tsinghua University,Z.ai (Zhipu AI)", "orgCategory": "Academia,Industry", "country": "China", "countries": [ "China" ], "date": "2023-12-14", "year": 2023, "decimalYear": 2023.950685, "domains": [ "Vision", "Language" ], "task": "Instruction interpretation,Visual question answering", "parameters": 18000000000, "computeFlop": 6.707e+22, "log10Compute": 22.827, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Likely", "link": "https://arxiv.org/abs/2312.08914" }, { "name": "nekomata-14b", "organization": "rinna", "orgCategory": "Industry", "country": "Japan", "countries": [ "Japan" ], "date": "2023-12-21", "year": 2023, "decimalYear": 2023.969863, "domains": [ "Language" ], "task": "Language generation", "parameters": 14200000000, "computeFlop": 2.5562e+23, "log10Compute": 23.408, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://arxiv.org/abs/2404.01657" }, { "name": "GQA-8-XXL", "organization": "Google Research", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2023-12-23", "year": 2023, "decimalYear": 2023.975342, "domains": [ "Language" ], "task": "Text summarization,Language modeling/generation,Translation", "parameters": 11000000000, "computeFlop": 3.4912896e+22, "log10Compute": 22.543, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2305.13245" }, { "name": "Qwen1.5-72B", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2024-02-04", "year": 2024, "decimalYear": 2024.092896, "domains": [ "Language" ], "task": "Chat,Language modeling/generation,Quantitative reasoning,Code generation,Translation", "parameters": 72000000000, "computeFlop": 1.3e+24, "log10Compute": 24.114, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://qwenlm.github.io/blog/qwen1.5/" }, { "name": "Stable Diffusion 3", "organization": "Stability AI", "orgCategory": "Industry", "country": "United Kingdom", "countries": [ "United Kingdom" ], "date": "2024-02-22", "year": 2024, "decimalYear": 2024.142077, "domains": [ "Image generation" ], "task": "Image generation,Text-to-image", "parameters": 8000000000, "computeFlop": 5.0000000000000004e+22, "log10Compute": 22.699, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Confident", "link": "https://arxiv.org/abs/2403.03206\nhttps://stability.ai/news/stable-diffusion-3" }, { "name": "MegaScale (Production)", "organization": "ByteDance,Peking University", "orgCategory": "Industry,Academia", "country": "China", "countries": [ "China" ], "date": "2024-02-23", "year": 2024, "decimalYear": 2024.144809, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 530000000000, "computeFlop": 3.9e+24, "log10Compute": 24.591, "trainingCostUsd2023": 2614019.245, "powerDrawW": 9728028.18454986, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Speculative", "link": "https://arxiv.org/abs/2402.15627" }, { "name": "Mistral Large", "organization": "Mistral AI", "orgCategory": "Industry", "country": "France", "countries": [ "France" ], "date": "2024-02-26", "year": 2024, "decimalYear": 2024.153005, "domains": [ "Language" ], "task": "Chat", "parameters": null, "computeFlop": 1.12e+25, "log10Compute": 25.049, "trainingCostUsd2023": 14110111.93, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Likely", "link": "https://mistral.ai/news/mistral-large/" }, { "name": "Aramco Metabrain AI", "organization": "Saudi Aramco", "orgCategory": "Industry,Government", "country": "Saudi Arabia", "countries": [ "Saudi Arabia" ], "date": "2024-03-04", "year": 2024, "decimalYear": 2024.172131, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 250000000000, "computeFlop": 1.05e+25, "log10Compute": 25.021, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://www.offshore-technology.com/news/saudi-aramco-unveils-industry-first-generative-ai-model/" }, { "name": "Inflection-2.5", "organization": "Inflection AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-03-07", "year": 2024, "decimalYear": 2024.180328, "domains": [ "Language" ], "task": "Chat", "parameters": null, "computeFlop": 8.000001e+24, "log10Compute": 24.903, "trainingCostUsd2023": 11804593.33, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Hosted access (no API)", "confidence": "Likely", "link": "https://inflection.ai/inflection-2-5" }, { "name": "MM1-30B", "organization": "Apple", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-03-14", "year": 2024, "decimalYear": 2024.199454, "domains": [ "Multimodal", "Language", "Vision" ], "task": "Chat,Image captioning,Visual question answering", "parameters": 30000000000, "computeFlop": 4.86e+23, "log10Compute": 23.687, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2403.09611" }, { "name": "DBRX", "organization": "Databricks", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-03-27", "year": 2024, "decimalYear": 2024.234973, "domains": [ "Language" ], "task": "Chat,Code generation", "parameters": 132000000000, "computeFlop": 2.6e+24, "log10Compute": 24.415, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://www.databricks.com/blog/introducing-dbrx-new-state-art-open-llm" }, { "name": "Reka Core", "organization": "Reka AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-04-15", "year": 2024, "decimalYear": 2024.286885, "domains": [ "Multimodal", "Language", "Vision", "Video", "Speech" ], "task": "Chat,Language modeling/generation,Image captioning,Code generation,Code autocompletion,Question answering,Visual question answering,Video description,Speech recognition (ASR),Speech-to-text,Quantitative reasoning", "parameters": 67000000000, "computeFlop": 8.400010000000001e+24, "log10Compute": 24.924, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Speculative", "link": "https://arxiv.org/abs/2404.12387" }, { "name": "Llama 3-70B", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-04-18", "year": 2024, "decimalYear": 2024.295082, "domains": [ "Language" ], "task": "Chat,Language modeling/generation,Code generation", "parameters": 70000000000, "computeFlop": 7.861e+24, "log10Compute": 24.895, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://ai.meta.com/blog/meta-llama-3/" }, { "name": "GenCast", "organization": "Google DeepMind", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-05-01", "year": 2024, "decimalYear": 2024.330601, "domains": [ "Earth science" ], "task": "Weather forecasting", "parameters": null, "computeFlop": 816998400000000000000, "log10Compute": 20.912, "trainingCostUsd2023": null, "powerDrawW": 14228.47853322057, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2312.15796" }, { "name": "VILA1.5-13B", "organization": "NVIDIA,Massachusetts Institute of Technology (MIT)", "orgCategory": "Industry,Academia", "country": "United States", "countries": [ "United States" ], "date": "2024-05-03", "year": 2024, "decimalYear": 2024.336066, "domains": [ "Multimodal", "Language", "Vision", "Video" ], "task": "Chat,Visual question answering,Image captioning,Language modeling/generation,Question answering", "parameters": 13493916736, "computeFlop": 2.3003136e+21, "log10Compute": 21.362, "trainingCostUsd2023": null, "powerDrawW": 101175.78544917757, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://huggingface.co/Efficient-Large-Model/VILA1.5-13b\nhttps://github.com/NVlabs/VILA/tree/bbc609baf326b1b49b93450b48edc516db3737fc/scripts/v1_5/release/13b\nhttps://developer.nvidia.com/blog/visual-language-models-on-nvidia-hardware-with-vila/" }, { "name": "AlphaFold 3", "organization": "Google DeepMind,Isomorphic Labs", "orgCategory": "Industry,Industry", "country": "United States, United Kingdom", "countries": [ "United States", "United Kingdom" ], "date": "2024-05-08", "year": 2024, "decimalYear": 2024.349727, "domains": [ "Biology" ], "task": "Protein folding prediction,Antibody property prediction,Protein-ligand contact prediction,RNA structure prediction,Protein interaction prediction", "parameters": null, "computeFlop": 4.1405645e+22, "log10Compute": 22.617, "trainingCostUsd2023": null, "powerDrawW": 202329.0409413959, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://www.nature.com/articles/s41586-024-07487-w" }, { "name": "Yi-Large", "organization": "01.AI", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2024-05-13", "year": 2024, "decimalYear": 2024.363388, "domains": [ "Language" ], "task": "Chat,Language modeling/generation", "parameters": 100000000000, "computeFlop": 1.8e+24, "log10Compute": 24.255, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Speculative", "link": null }, { "name": "Octo-Base", "organization": "University of California (UC) Berkeley,Stanford University,Carnegie Mellon University (CMU),DeepMind", "orgCategory": "Academia,Academia,Academia,Industry", "country": "United States, United Kingdom", "countries": [ "United States", "United Kingdom" ], "date": "2024-05-20", "year": 2024, "decimalYear": 2024.382514, "domains": [ "Robotics" ], "task": "Robotic manipulation", "parameters": 93000000, "computeFlop": 585000000000000000000, "log10Compute": 20.767, "trainingCostUsd2023": null, "powerDrawW": 85966.86619350823, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2405.12213" }, { "name": "ALLaM adapted 70B", "organization": "Saudi Data and Artificial Intelligence Authority", "orgCategory": "Industry,Government", "country": "Saudi Arabia", "countries": [ "Saudi Arabia" ], "date": "2024-05-21", "year": 2024, "decimalYear": 2024.385246, "domains": [ "Language" ], "task": "Language modeling/generation,Translation,Question answering", "parameters": 70000000000, "computeFlop": 1.062e+24, "log10Compute": 24.026, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2407.15390" }, { "name": "Qwen2-72B", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2024-06-07", "year": 2024, "decimalYear": 2024.431694, "domains": [ "Language" ], "task": "Chat,Language modeling/generation,Question answering", "parameters": 72710000000, "computeFlop": 3.02e+24, "log10Compute": 24.48, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://qwenlm.github.io/blog/qwen2/ \nhttps://arxiv.org/abs/2407.10671" }, { "name": "Llama-3.1-Nemotron-70B-Instruct", "organization": "NVIDIA,Meta AI", "orgCategory": "Industry,Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-06-12", "year": 2024, "decimalYear": 2024.445355, "domains": [ "Language" ], "task": "Language modeling", "parameters": null, "computeFlop": 7.929e+24, "log10Compute": 24.899, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://catalog.ngc.nvidia.com/orgs/nim/teams/nvidia/containers/llama-3.1-nemotron-70b-instruct" }, { "name": "OpenVLA", "organization": "Stanford University,University of California (UC) Berkeley,Toyota Research Institute,Google DeepMind,Massachusetts Institute of Technology (MIT),Physical Intelligence", "orgCategory": "Academia,Academia,Industry,Industry,Academia,Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-06-13", "year": 2024, "decimalYear": 2024.448087, "domains": [ "Robotics", "Vision", "Language" ], "task": "Robotic manipulation", "parameters": 7188100000, "computeFlop": 1.1e+23, "log10Compute": 23.041, "trainingCostUsd2023": null, "powerDrawW": 50541.724821294745, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://openvla.github.io/\nhttps://arxiv.org/abs/2406.09246" }, { "name": "Nemotron-4 340B", "organization": "NVIDIA", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-06-14", "year": 2024, "decimalYear": 2024.45082, "domains": [ "Language" ], "task": "Language modeling/generation,Chat,Question answering", "parameters": 340000000000, "computeFlop": 1.8e+25, "log10Compute": 25.255, "trainingCostUsd2023": 21271017.96222655, "powerDrawW": 8490820.682633882, "frontier": true, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://blogs.nvidia.com/blog/nemotron-4-synthetic-data-generation-llm-training/" }, { "name": "DeepSeek-Coder-V2 236B", "organization": "DeepSeek", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2024-06-17", "year": 2024, "decimalYear": 2024.459016, "domains": [ "Language" ], "task": "Code generation,Code autocompletion", "parameters": 236000000000, "computeFlop": 1.2852e+24, "log10Compute": 24.109, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://github.com/deepseek-ai/DeepSeek-Coder-V2" }, { "name": "Claude 3.5 Sonnet", "organization": "Anthropic", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-06-20", "year": 2024, "decimalYear": 2024.467213, "domains": [ "Multimodal", "Language", "Vision" ], "task": "Chat,Image captioning,Code generation,Language modeling/generation", "parameters": null, "computeFlop": 2.700000000000001e+25, "log10Compute": 25.431, "trainingCostUsd2023": 25870993.12024943, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Speculative", "link": "https://www-cdn.anthropic.com/fed9cc193a14b84131812372d8d5857f8f304c52/Model_Card_Claude_3_Addendum.pdf" }, { "name": "ESM3 (98B)", "organization": "EvolutionaryScale,University of California (UC) Berkeley", "orgCategory": "Industry,Academia", "country": "United States", "countries": [ "United States" ], "date": "2024-06-25", "year": 2024, "decimalYear": 2024.480874, "domains": [ "Biology" ], "task": "Protein generation", "parameters": 98500000000, "computeFlop": 1.07e+24, "log10Compute": 24.029, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://www.evolutionaryscale.ai/blog/esm3-release" }, { "name": "Llama 3.1-405B", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-07-23", "year": 2024, "decimalYear": 2024.557377, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Code generation,Mathematical reasoning", "parameters": 405000000000, "computeFlop": 3.8e+25, "log10Compute": 25.58, "trainingCostUsd2023": 52885433.95402246, "powerDrawW": 22622532.159299146, "frontier": true, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://ai.meta.com/research/publications/the-llama-3-herd-of-models/" }, { "name": "Mistral Large 2", "organization": "Mistral AI", "orgCategory": "Industry", "country": "France", "countries": [ "France" ], "date": "2024-07-24", "year": 2024, "decimalYear": 2024.560109, "domains": [ "Language" ], "task": "Language modeling/generation,Translation,Code generation", "parameters": 123000000000, "computeFlop": 2.13e+25, "log10Compute": 25.328, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Likely", "link": "https://mistral.ai/news/mistral-large-2407/" }, { "name": "AFM-server", "organization": "Apple", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-07-29", "year": 2024, "decimalYear": 2024.57377, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": null, "computeFlop": 4.3e+24, "log10Compute": 24.633, "trainingCostUsd2023": null, "powerDrawW": 5493309.4800721435, "frontier": false, "openWeights": false, "accessibility": "Hosted access (no API)", "confidence": "Likely", "link": "https://machinelearning.apple.com/research/apple-intelligence-foundation-language-models" }, { "name": "AFM-on-device", "organization": "Apple", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-07-29", "year": 2024, "decimalYear": 2024.57377, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 2730000000, "computeFlop": 4.5126e+23, "log10Compute": 23.654, "trainingCostUsd2023": null, "powerDrawW": 2181166.9994404097, "frontier": false, "openWeights": false, "accessibility": "Hosted access (no API)", "confidence": "Confident", "link": "https://machinelearning.apple.com/research/apple-intelligence-foundation-language-models" }, { "name": "LLaVA-OV-72B", "organization": "ByteDance,Nanyang Technological University,Chinese University of Hong Kong (CUHK),Hong Kong University of Science and Technology (HKUST)", "orgCategory": "Industry,Academia,Academia,Academia", "country": "China, Singapore, Hong Kong", "countries": [ "China", "Singapore", "Hong Kong" ], "date": "2024-08-06", "year": 2024, "decimalYear": 2024.595628, "domains": [ "Multimodal", "Vision", "Language", "Video" ], "task": "Image captioning,Visual question answering,Video description,Object recognition,Action recognition,Language modeling/generation", "parameters": 72000000000, "computeFlop": 3.036551985824e+24, "log10Compute": 24.482, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2408.03326" }, { "name": "Grok-2", "organization": "xAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-08-13", "year": 2024, "decimalYear": 2024.614754, "domains": [ "Language", "Vision", "Multimodal" ], "task": "Chat,Language modeling/generation,Question answering,Code generation,Visual question answering", "parameters": null, "computeFlop": 2.96e+25, "log10Compute": 25.471, "trainingCostUsd2023": 31602310.530835517, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Confident", "link": "https://x.ai/blog/grok-2" }, { "name": "DeepSeek-V2.5", "organization": "DeepSeek", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2024-09-06", "year": 2024, "decimalYear": 2024.680328, "domains": [ "Language" ], "task": "Language modeling/generation,Chat,Code generation", "parameters": 236000000000, "computeFlop": 1.7892e+24, "log10Compute": 24.253, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://huggingface.co/deepseek-ai/DeepSeek-V2.5" }, { "name": "Qwen2.5-32B", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2024-09-17", "year": 2024, "decimalYear": 2024.710383, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Quantitative reasoning", "parameters": 32500000000, "computeFlop": 3.51e+24, "log10Compute": 24.545, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://qwenlm.github.io/blog/qwen2.5/" }, { "name": "Qwen2.5-72B", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2024-09-19", "year": 2024, "decimalYear": 2024.715847, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Quantitative reasoning", "parameters": 72700000000, "computeFlop": 7.8e+24, "log10Compute": 24.892, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://qwenlm.github.io/blog/qwen2.5/" }, { "name": "Qwen2.5 Instruct (72B)", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2024-09-19", "year": 2024, "decimalYear": 2024.715847, "domains": [ "Language" ], "task": "Code generation,Code autocompletion,Quantitative reasoning,Question answering,Language modeling/generation", "parameters": 72700000000, "computeFlop": 7.8516e+24, "log10Compute": 24.895, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://qwenlm.github.io/blog/qwen2.5/" }, { "name": "Telechat2-115B", "organization": "China Telecom", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2024-09-20", "year": 2024, "decimalYear": 2024.718579, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Quantitative reasoning,Chat,Text summarization,Code generation", "parameters": 115000000000, "computeFlop": 6.9e+24, "log10Compute": 24.839, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://huggingface.co/Tele-AI/TeleChat2-115B" }, { "name": "Llama 3.2 11B", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-09-24", "year": 2024, "decimalYear": 2024.729508, "domains": [ "Multimodal", "Vision", "Language" ], "task": "Visual question answering,Image captioning,Object detection", "parameters": 10600000000, "computeFlop": 5.79e+23, "log10Compute": 23.763, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://ai.meta.com/blog/llama-3-2-connect-2024-vision-edge-mobile-devices/" }, { "name": "Movie Gen Video", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-10-04", "year": 2024, "decimalYear": 2024.756831, "domains": [ "Video", "Vision" ], "task": "Video generation,Text-to-video,Image-to-video", "parameters": 30000000000, "computeFlop": 1.65e+24, "log10Compute": 24.217, "trainingCostUsd2023": null, "powerDrawW": 8469669.524206791, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://ai.meta.com/static-resource/movie-gen-research-paper" }, { "name": "RDT-1B", "organization": "Tsinghua University", "orgCategory": "Academia", "country": "China", "countries": [ "China" ], "date": "2024-10-10", "year": 2024, "decimalYear": 2024.773224, "domains": [ "Robotics" ], "task": "Robotic manipulation", "parameters": 1200000000, "computeFlop": 4.06e+22, "log10Compute": 22.609, "trainingCostUsd2023": null, "powerDrawW": 33080.22622858553, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2410.07864" }, { "name": "CHAI-1", "organization": "Chai discovery", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-10-15", "year": 2024, "decimalYear": 2024.786885, "domains": [ "Biology" ], "task": "Protein folding prediction,Protein-ligand contact prediction", "parameters": null, "computeFlop": 7.7605724e+21, "log10Compute": 21.89, "trainingCostUsd2023": null, "powerDrawW": 100804.70264211958, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://www.chaidiscovery.com/blog/introducing-chai-1\nhttps://www.biorxiv.org/content/10.1101/2024.10.10.615955v2" }, { "name": "Yi-Lightning", "organization": "01.AI", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2024-10-18", "year": 2024, "decimalYear": 2024.795082, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": null, "computeFlop": 1.5e+24, "log10Compute": 24.176, "trainingCostUsd2023": null, "powerDrawW": 2756194.445567179, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Confident", "link": "https://www.lingyiwanwu.com/en https://platform.lingyiwanwu.com/" }, { "name": "NVLM-X 72B", "organization": "NVIDIA", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-10-22", "year": 2024, "decimalYear": 2024.806011, "domains": [ "Vision", "Language" ], "task": "Language modeling/generation,Vision-language generation,Question answering,Code generation,Translation,Quantitative reasoning", "parameters": 72000000000, "computeFlop": 3.0398181e+24, "log10Compute": 24.483, "trainingCostUsd2023": null, "powerDrawW": 176380.73226445544, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2409.11402" }, { "name": "NVLM-H 72B", "organization": "NVIDIA", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-10-22", "year": 2024, "decimalYear": 2024.806011, "domains": [ "Vision", "Language" ], "task": "Language modeling/generation,Vision-language generation,Question answering,Code generation,Translation,Quantitative reasoning", "parameters": 72000000000, "computeFlop": 3.02e+24, "log10Compute": 24.48, "trainingCostUsd2023": null, "powerDrawW": 176380.73226445544, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://arxiv.org/abs/2409.11402" }, { "name": "NVLM-D 72B", "organization": "NVIDIA", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-10-22", "year": 2024, "decimalYear": 2024.806011, "domains": [ "Vision", "Language" ], "task": "Language modeling/generation,Vision-language generation,Question answering,Code generation,Translation,Quantitative reasoning", "parameters": 72000000000, "computeFlop": 3.02e+24, "log10Compute": 24.48, "trainingCostUsd2023": null, "powerDrawW": 176380.73226445544, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2409.11402" }, { "name": "Doubao-pro", "organization": "ByteDance", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2024-10-28", "year": 2024, "decimalYear": 2024.822404, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Text summarization,Text classification", "parameters": 500000000000, "computeFlop": 2.505e+25, "log10Compute": 25.399, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Likely", "link": "https://www.volcengine.com/docs/6360/1264663" }, { "name": "Hunyuan-Large", "organization": "Tencent", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2024-11-06", "year": 2024, "decimalYear": 2024.846995, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Code generation,Translation", "parameters": 389000000000, "computeFlop": 3.49237e+24, "log10Compute": 24.543, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://arxiv.org/abs/2411.02265" }, { "name": "Amazon Nova Pro", "organization": "Amazon", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-12-03", "year": 2024, "decimalYear": 2024.920765, "domains": [ "Multimodal", "Language", "Video", "Vision" ], "task": "Language modeling/generation,Retrieval-augmented generation,Visual question answering,Image captioning,Video description,Character recognition (OCR),Code generation,Translation", "parameters": null, "computeFlop": 6.000010000000001e+24, "log10Compute": 24.778, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Speculative", "link": "https://aws.amazon.com/es/blogs/aws/introducing-amazon-nova-frontier-intelligence-and-industry-leading-price-performance/\n\nhttps://assets.amazon.science/96/7d/0d3e59514abf8fdcfafcdc574300/nova-tech-report-20250317-0810.pdf" }, { "name": "Llama 3.3 70B", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2024-12-06", "year": 2024, "decimalYear": 2024.928962, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Translation,Code generation", "parameters": 70000000000, "computeFlop": 6.8649768e+24, "log10Compute": 24.837, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://www.llama.com/docs/model-cards-and-prompt-formats/llama3_3/" }, { "name": "EXAONE 3.5 32B", "organization": "LG AI Research", "orgCategory": "Industry", "country": "South Korea", "countries": [ "South Korea" ], "date": "2024-12-09", "year": 2024, "decimalYear": 2024.937158, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Translation", "parameters": 32000000000, "computeFlop": 1.25e+24, "log10Compute": 24.097, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2412.04862" }, { "name": "DeepSeek-V3", "organization": "DeepSeek", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2024-12-24", "year": 2024, "decimalYear": 2024.978142, "domains": [ "Language" ], "task": "Language modeling/generation,Code generation,Quantitative reasoning,Question answering", "parameters": 671000000000, "computeFlop": 3.3e+24, "log10Compute": 24.519, "trainingCostUsd2023": 5390000, "powerDrawW": 2818135.1812142837, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://arxiv.org/abs/2412.19437" }, { "name": "Eagle 2", "organization": "NVIDIA,Nanjing University,Tsinghua University,Hong Kong Polytechnic University,Johns Hopkins University,New York University (NYU)", "orgCategory": "Industry,Academia,Academia,Academia,Academia,Academia", "country": "United States, China, Hong Kong", "countries": [ "United States", "China", "Hong Kong" ], "date": "2025-01-20", "year": 2025, "decimalYear": 2025.052055, "domains": [ "Vision", "Robotics", "Language" ], "task": null, "parameters": 8930000000, "computeFlop": 4.7156e+22, "log10Compute": 22.674, "trainingCostUsd2023": null, "powerDrawW": 352055.152733459, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "arxiv.org/abs/2501.14818" }, { "name": "DeepSeek-R1", "organization": "DeepSeek", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-01-20", "year": 2025, "decimalYear": 2025.052055, "domains": [ "Language" ], "task": "Language modeling/generation,Code generation,Quantitative reasoning,Question answering", "parameters": 671000000000, "computeFlop": 3.5e+24, "log10Compute": 24.544, "trainingCostUsd2023": 6770000, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://api-docs.deepseek.com/news/news250120" }, { "name": "Grok 3", "organization": "xAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2025-02-17", "year": 2025, "decimalYear": 2025.128767, "domains": [ "Language", "Vision", "Multimodal" ], "task": "Chat,Language modeling/generation,Question answering,Code generation,Visual question answering", "parameters": 3000000000000, "computeFlop": 3.5e+26, "log10Compute": 26.544, "trainingCostUsd2023": 217835545.5267339, "powerDrawW": 109948656.20196915, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Likely", "link": "https://x.ai/blog/grok-3" }, { "name": "Claude 3.7 Sonnet", "organization": "Anthropic", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2025-02-24", "year": 2025, "decimalYear": 2025.147945, "domains": [ "Language", "Vision", "Multimodal" ], "task": "Language modeling/generation,Question answering,Code generation,Quantitative reasoning,Translation,Instruction interpretation,Visual question answering", "parameters": null, "computeFlop": 3.35e+25, "log10Compute": 25.525, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Likely", "link": "https://www.anthropic.com/news/claude-3-7-sonnet" }, { "name": "GPT-4.5", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2025-02-27", "year": 2025, "decimalYear": 2025.156164, "domains": [ "Language", "Vision", "Multimodal" ], "task": "Language modeling/generation,Question answering,Quantitative reasoning,Translation,Visual question answering,Code generation,Instruction interpretation", "parameters": null, "computeFlop": 3.8e+26, "log10Compute": 26.58, "trainingCostUsd2023": 366010627.6281514, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Likely", "link": "https://openai.com/index/introducing-gpt-4-5/" }, { "name": "QwQ-32B", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-03-06", "year": 2025, "decimalYear": 2025.175342, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Quantitative reasoning,Code generation", "parameters": 32500000000, "computeFlop": 3.51e+24, "log10Compute": 24.545, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Speculative", "link": "https://qwenlm.github.io/blog/qwq-32b/" }, { "name": "Hunyuan-TurboS", "organization": "Tencent", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-03-11", "year": 2025, "decimalYear": 2025.189041, "domains": [ "Language" ], "task": "Language modeling/generation,Quantitative reasoning,Question answering,Code generation,Text summarization", "parameters": 560000000000, "computeFlop": 5.4e+24, "log10Compute": 24.732, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Confident", "link": "https://web.archive.org/web/20250408105622/https://www.dapingtime.com/article/2171.html\n\nhttps://medium.com/data-science-in-your-pocket/tencent-hunyuan-turbo-s-the-fastest-reasoning-llm-d64a02bed5c8\n\nhttps://arxiv.org/abs/2505.15431" }, { "name": "EXAONE Deep 32B", "organization": "LG AI Research", "orgCategory": "Industry", "country": "South Korea", "countries": [ "South Korea" ], "date": "2025-03-16", "year": 2025, "decimalYear": 2025.20274, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Quantitative reasoning,Code generation", "parameters": 32000000000, "computeFlop": 1.26e+24, "log10Compute": 24.1, "trainingCostUsd2023": null, "powerDrawW": 703248.4282353076, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2503.12524" }, { "name": "DeepSeek-V3 (Mar 2025)", "organization": "DeepSeek", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-03-24", "year": 2025, "decimalYear": 2025.224658, "domains": [ "Language" ], "task": "Language modeling/generation,Code generation,Quantitative reasoning,Question answering", "parameters": 671000000000, "computeFlop": 3.3e+24, "log10Compute": 24.519, "trainingCostUsd2023": 5390000, "powerDrawW": 2812492.608768545, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://arxiv.org/abs/2412.19437" }, { "name": "Llama 4 Scout", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2025-04-05", "year": 2025, "decimalYear": 2025.257534, "domains": [ "Multimodal", "Language", "Vision" ], "task": "Chat,Code generation,Visual question answering,Language modeling/generation,Question answering", "parameters": 109000000000, "computeFlop": 4.08e+24, "log10Compute": 24.611, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Likely", "link": "https://ai.meta.com/blog/llama-4-multimodal-intelligence/" }, { "name": "Llama 4 Maverick", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2025-04-05", "year": 2025, "decimalYear": 2025.257534, "domains": [ "Multimodal", "Language", "Vision" ], "task": "Chat,Code generation,Visual question answering,Language modeling/generation,Question answering", "parameters": 400000000000, "computeFlop": 2.244000000001e+24, "log10Compute": 24.351, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Likely", "link": "https://ai.meta.com/blog/llama-4-multimodal-intelligence/" }, { "name": "Llama 4 Behemoth (preview)", "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2025-04-05", "year": 2025, "decimalYear": 2025.257534, "domains": [ "Multimodal", "Language", "Vision" ], "task": "Chat,Code generation,Visual question answering,Translation,Language modeling/generation,Quantitative reasoning,Question answering", "parameters": 2000000000000, "computeFlop": 5.18400000000001e+25, "log10Compute": 25.715, "trainingCostUsd2023": 44588963.624007165, "powerDrawW": 43933454.99810563, "frontier": true, "openWeights": false, "accessibility": "Unreleased", "confidence": "Likely", "link": "https://ai.meta.com/blog/llama-4-multimodal-intelligence/" }, { "name": "Pangu Ultra", "organization": "Huawei", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-04-10", "year": 2025, "decimalYear": 2025.271233, "domains": [ "Language" ], "task": "Code generation,Language modeling/generation", "parameters": 135000000000, "computeFlop": 1.0692e+25, "log10Compute": 25.029, "trainingCostUsd2023": null, "powerDrawW": 6426121.277196856, "frontier": false, "openWeights": false, "accessibility": "Hosted access (no API)", "confidence": "Confident", "link": "https://arxiv.org/abs/2504.07866" }, { "name": "Qwen3-235B-A22B", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-04-29", "year": 2025, "decimalYear": 2025.323288, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Mathematical reasoning,Quantitative reasoning,Code generation,Translation", "parameters": 235000000000, "computeFlop": 4.752e+24, "log10Compute": 24.677, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://qwenlm.github.io/blog/qwen3/" }, { "name": "Seed1.5-VL", "organization": "ByteDance", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-05-11", "year": 2025, "decimalYear": 2025.356164, "domains": [ "Vision", "Language", "Multimodal", "Video" ], "task": "Visual question answering,Video description,Language modeling/generation,Question answering,Character recognition (OCR)", "parameters": null, "computeFlop": 1.388556e+24, "log10Compute": 24.143, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Confident", "link": "https://arxiv.org/abs/2505.07062" }, { "name": "DeepSeek-R1 (May 2025)", "organization": "DeepSeek", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-05-28", "year": 2025, "decimalYear": 2025.40274, "domains": [ "Language" ], "task": "Language modeling/generation,Code generation,Quantitative reasoning,Question answering", "parameters": 671000000000, "computeFlop": 4.020010000000001e+24, "log10Compute": 24.604, "trainingCostUsd2023": 6770000, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://api-docs.deepseek.com/news/news250120" }, { "name": "FGN", "organization": "Google DeepMind", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2025-06-12", "year": 2025, "decimalYear": 2025.443836, "domains": [ "Earth science" ], "task": "Weather forecasting", "parameters": 720000000, "computeFlop": 9.618950880000001e+21, "log10Compute": 21.983, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "Unreleased", "confidence": "Confident", "link": "https://arxiv.org/abs/2506.10772v1" }, { "name": "Grok 4", "organization": "xAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2025-07-09", "year": 2025, "decimalYear": 2025.517808, "domains": [ "Language", "Multimodal", "Vision" ], "task": "Language modeling/generation,Question answering,Search,Visual question answering,Character recognition (OCR),Image captioning,Quantitative reasoning", "parameters": 3000000000000, "computeFlop": 5.0000000000001e+26, "log10Compute": 26.699, "trainingCostUsd2023": 387842678.08636147, "powerDrawW": null, "frontier": true, "openWeights": false, "accessibility": "API access", "confidence": "Speculative", "link": "https://x.ai/news/grok-4" }, { "name": "Kimi K2", "organization": "Moonshot", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-07-11", "year": 2025, "decimalYear": 2025.523288, "domains": [ "Language" ], "task": "Language modeling/generation,Code generation,Question answering,Quantitative reasoning,Search,Table tasks", "parameters": 1000000000000, "computeFlop": 2.976e+24, "log10Compute": 24.474, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "https://moonshotai.github.io/Kimi-K2/" }, { "name": "EXAONE 4.0 (32B)", "organization": "LG AI Research", "orgCategory": "Industry", "country": "South Korea", "countries": [ "South Korea" ], "date": "2025-07-15", "year": 2025, "decimalYear": 2025.534247, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Code generation,Quantitative reasoning,Translation", "parameters": 32000000000, "computeFlop": 2.69000000000001e+24, "log10Compute": 24.43, "trainingCostUsd2023": null, "powerDrawW": 701356.0100875632, "frontier": false, "openWeights": true, "accessibility": "Open weights (non-commercial)", "confidence": "Confident", "link": "https://arxiv.org/abs/2507.11407" }, { "name": "Qwen3-Coder-480B-A35B", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-07-22", "year": 2025, "decimalYear": 2025.553425, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Code generation,System control", "parameters": 480000000000, "computeFlop": 1.575e+24, "log10Compute": 24.197, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://qwenlm.github.io/blog/qwen3-coder/" }, { "name": "Qwen3-235B-A22B-Thinking (Jul 2025)", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-07-25", "year": 2025, "decimalYear": 2025.561644, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Mathematical reasoning,Quantitative reasoning,Code generation,Translation", "parameters": 235000000000, "computeFlop": 4.752e+24, "log10Compute": 24.677, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://qwenlm.github.io/blog/qwen3/" }, { "name": "Qwen3-235B-A22B (Jul 2025)", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-07-25", "year": 2025, "decimalYear": 2025.561644, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Mathematical reasoning,Quantitative reasoning,Code generation,Translation", "parameters": 235000000000, "computeFlop": 4.752e+24, "log10Compute": 24.677, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://qwenlm.github.io/blog/qwen3/" }, { "name": "gpt-oss-120b", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2025-08-05", "year": 2025, "decimalYear": 2025.591781, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering", "parameters": 116830000000, "computeFlop": 4.94e+24, "log10Compute": 24.694, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://cdn.openai.com/pdf/419b6906-9da6-406c-a19d-1bb078ac7637/oai_gpt-oss_model_card.pdf" }, { "name": "gpt-oss-20b", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2025-08-05", "year": 2025, "decimalYear": 2025.591781, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering", "parameters": 20910000000, "computeFlop": 5.49e+23, "log10Compute": 23.74, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://cdn.openai.com/pdf/419b6906-9da6-406c-a19d-1bb078ac7637/oai_gpt-oss_model_card.pdf" }, { "name": "GLM-4.5", "organization": "Z.ai (Zhipu AI),Tsinghua University", "orgCategory": "Industry,Academia", "country": "China", "countries": [ "China" ], "date": "2025-08-05", "year": 2025, "decimalYear": 2025.591781, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Code generation,Quantitative reasoning", "parameters": 355000000000, "computeFlop": 4.42e+24, "log10Compute": 24.645, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2508.06471" }, { "name": "GPT-5", "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2025-08-07", "year": 2025, "decimalYear": 2025.59726, "domains": [ "Multimodal", "Language", "Vision" ], "task": "Language modeling/generation,Question answering", "parameters": null, "computeFlop": 6.6e+25, "log10Compute": 25.82, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Speculative", "link": "https://cdn.openai.com/gpt-5-system-card.pdf\nhttps://openai.com/index/introducing-gpt-5/" }, { "name": "LongCat-Flash", "organization": "Meituan Inc", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-09-01", "year": 2025, "decimalYear": 2025.665753, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Chat,Code generation,Quantitative reasoning,Instruction interpretation", "parameters": 560000000000, "computeFlop": 3.726e+24, "log10Compute": 24.571, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2509.01322" }, { "name": "Qwen3-Max", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-09-05", "year": 2025, "decimalYear": 2025.676712, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Mathematical reasoning,Code generation,Quantitative reasoning,Retrieval-augmented generation,Translation", "parameters": 1000000000000, "computeFlop": 1.512e+25, "log10Compute": 25.18, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Speculative", "link": "https://modelstudio.console.alibabacloud.com/?tab=doc#/doc/?type=model&url=2840914_2&modelId=qwen3-max-preview\n\nhttps://qwen.ai/blog?id=87dc93fc8a590dc718c77e1f6e84c07b474f6c5a&from=home.latest-research-list" }, { "name": "AgentFounder-30B", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-09-16", "year": 2025, "decimalYear": 2025.706849, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,System control,Quantitative reasoning,Mathematical reasoning,Code generation,Search", "parameters": 30000000000, "computeFlop": 6.5367e+23, "log10Compute": 23.815, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2509.13310" }, { "name": "Qwen3-Omni-30B-A3B", "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-09-22", "year": 2025, "decimalYear": 2025.723288, "domains": [ "Multimodal", "Language", "Vision", "Speech", "Video" ], "task": "Language modeling/generation,Question answering,Visual question answering,Image captioning,Video description,Speech recognition (ASR),Speech synthesis,Speech-to-text,Text-to-speech (TTS)", "parameters": 35300000000, "computeFlop": 3.6e+22, "log10Compute": 22.556, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2509.17765" }, { "name": "GLM-4.6", "organization": "Z.ai (Zhipu AI),Tsinghua University", "orgCategory": "Industry,Academia", "country": "China", "countries": [ "China" ], "date": "2025-09-30", "year": 2025, "decimalYear": 2025.745205, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Code generation,Quantitative reasoning,System control", "parameters": 357000000000, "computeFlop": 4.42e+24, "log10Compute": 24.645, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://z.ai/blog/glm-4.6" }, { "name": "Ling-1T", "organization": "Ant Group", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-10-10", "year": 2025, "decimalYear": 2025.772603, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Quantitative reasoning,Code generation,Mathematical reasoning", "parameters": 1000000000000, "computeFlop": 6.000001e+24, "log10Compute": 24.778, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://huggingface.co/inclusionAI/Ling-1T\nhttps://arxiv.org/pdf/2510.22115" }, { "name": "Kimi K2 Thinking", "organization": "Moonshot", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-11-06", "year": 2025, "decimalYear": 2025.846575, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering,Search,Quantitative reasoning,Code generation", "parameters": 1000000000000, "computeFlop": 4.2e+24, "log10Compute": 24.623, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Likely", "link": "https://moonshotai.github.io/Kimi-K2/thinking" }, { "name": "Olmo 3", "organization": "Allen Institute for AI", "orgCategory": "Research collective", "country": "United States", "countries": [ "United States" ], "date": "2025-11-20", "year": 2025, "decimalYear": 2025.884932, "domains": [ "Language" ], "task": "Language modeling/generation,Chat,Reasoning,Code generation", "parameters": 32000000000, "computeFlop": 1.1e+24, "log10Compute": 24.041, "trainingCostUsd2023": null, "powerDrawW": 1398719.3115551586, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Confident", "link": "https://arxiv.org/abs/2512.13961" }, { "name": "Nemotron 3-Nano-30B-A3B", "organization": "NVIDIA", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2025-12-15", "year": 2025, "decimalYear": 2025.953425, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 31600000000, "computeFlop": 4.8e+23, "log10Compute": 23.681, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Likely", "link": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3-Nano-Technical-Report.pdf" }, { "name": "GLM-4.7", "organization": "Z.ai (Zhipu AI)", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2025-12-22", "year": 2025, "decimalYear": 2025.972603, "domains": [ "Language" ], "task": "Language modeling/generation,Code generation", "parameters": 358000000000, "computeFlop": 4.42e+24, "log10Compute": 24.645, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://z.ai/blog/glm-4.7" }, { "name": "K-EXAONE", "organization": "LG AI Research", "orgCategory": "Industry", "country": "South Korea", "countries": [ "South Korea" ], "date": "2025-12-31", "year": 2025, "decimalYear": 2025.99726, "domains": [ "Language" ], "task": "Language modeling/generation,Chat,Question answering,Code generation,Mathematical reasoning,Instruction interpretation", "parameters": 236000000000, "computeFlop": 1.52e+24, "log10Compute": 24.182, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Confident", "link": "https://huggingface.co/LGAI-EXAONE" }, { "name": "Kimi K2.5", "organization": "Moonshot", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2026-02-02", "year": 2026, "decimalYear": 2026.087671, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 1040000000000, "computeFlop": 5.8e+24, "log10Compute": 24.763, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2602.02276" }, { "name": "GLM-5", "organization": "Z.ai (Zhipu AI)", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2026-02-17", "year": 2026, "decimalYear": 2026.128767, "domains": [ "Language" ], "task": "Language modeling/generation", "parameters": 744000000000, "computeFlop": 6.84e+24, "log10Compute": 24.835, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/pdf/2602.15763" }, { "name": "Composer 2", "organization": "Cursor", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2026-03-19", "year": 2026, "decimalYear": 2026.210959, "domains": [ "Language" ], "task": "Coding", "parameters": 1040000000000, "computeFlop": 2.32e+25, "log10Compute": 25.365, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": false, "accessibility": "API access", "confidence": "Confident", "link": "https://arxiv.org/abs/2603.24477" }, { "name": "EXAONE 4.5", "organization": "LG AI Research", "orgCategory": "Industry", "country": "South Korea", "countries": [ "South Korea" ], "date": "2026-04-09", "year": 2026, "decimalYear": 2026.268493, "domains": [ "Language", "Multimodal" ], "task": "Language modeling/generation,Question answering", "parameters": 33200000000, "computeFlop": 3.85e+24, "log10Compute": 24.585, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (restricted use)", "confidence": "Confident", "link": "Eunbi Choi, Kibong Choi, Sehyun Chun, Seokhee Hong, Junwon Hwang, Hyojin Jeon, Ahra Jo, Hyunjik Jo, Yeonsik Jo, Joonkee Kim, Seonghwan Kim, Soyeon Kim, Sunkyoung Kim, Yireun Kim, Yongil Kim, Changhun Lee, Haeju Lee, Jinsik Lee, Kyungmin Lee, Sangha Park, Kwangrok Ryoo, Minju Seo, Sejong Yang, Heuiyeen Yeen, Hwan Chang, Stanley Jungkyu Choi, Yejin Choi, Kyubeen Han, Joonwon Jang, Kijeong Jeon, Geunyeong Jeong, Gerrard Jeongwon Jo, Jiyeon Jung, Daeseong Kim, Dohoon Kim, Dohyun Kim, Hyunseo Kim, Minu Kim, Myoungshin Kim, Youchul Kim, Byungoh Ko, Christopher Lee, Edward Hwayoung Lee, Honglak Lee, Jiyoung Lee, Sangeun Lee, Seungwon Lim, Woohyung Lim, Jueun Mun, Jaewoo Park, Jimin Park, Jinho Park, Yongmin Park, Wooseok Seo, Yongwoo Song, Sihyuk Yi, Kyungjae Yoo, Sangyeon Yoon" }, { "name": "MiMo-V2.5-Pro", "organization": "Xiaomi Corp", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2026-04-23", "year": 2026, "decimalYear": 2026.306849, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering", "parameters": 1020000000000, "computeFlop": 6.804e+24, "log10Compute": 24.833, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://mimo.xiaomi.com/mimo-v2-5-pro/" }, { "name": "DeepSeek-V4-Pro", "organization": "DeepSeek", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2026-04-24", "year": 2026, "decimalYear": 2026.309589, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering", "parameters": 1600000000000, "computeFlop": 9.702e+24, "log10Compute": 24.987, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/blob/main/DeepSeek_V4.pdf" }, { "name": "DeepSeek-V4-Flash", "organization": "DeepSeek", "orgCategory": "Industry", "country": "China", "countries": [ "China" ], "date": "2026-04-24", "year": 2026, "decimalYear": 2026.309589, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering", "parameters": 284000000000, "computeFlop": 2.496e+24, "log10Compute": 24.397, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/blob/main/DeepSeek_V4.pdf" }, { "name": "Composer 2.5", "organization": "Cursor", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2026-05-18", "year": 2026, "decimalYear": 2026.375342, "domains": [ "Language" ], "task": "Coding", "parameters": 1040000000000, "computeFlop": 3.87e+25, "log10Compute": 25.588, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": null, "accessibility": null, "confidence": "Likely", "link": "https://cursor.com/blog/composer-2-5" }, { "name": "Nemotron 3 Ultra", "organization": "NVIDIA", "orgCategory": "Industry", "country": "United States", "countries": [ "United States" ], "date": "2026-06-04", "year": 2026, "decimalYear": 2026.421918, "domains": [ "Language" ], "task": "Language modeling/generation,Question answering", "parameters": 550000000000, "computeFlop": 6.6e+24, "log10Compute": 24.82, "trainingCostUsd2023": null, "powerDrawW": null, "frontier": false, "openWeights": true, "accessibility": "Open weights (unrestricted)", "confidence": "Likely", "link": "https://arxiv.org/abs/2512.20856\nhttps://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3-Ultra-Technical-Report.pdf" } ], "aggregates": { "computeTrend": { "overall": { "slopeLog10PerYear": 0.3275, "intercept": -640.6015, "r2": 0.7592, "doublingTimeMonths": 11, "startYear": 1950, "endYear": 2026, "n": 528 }, "modernEra": { "slopeLog10PerYear": 0.639, "intercept": -1270.1435, "r2": 0.6515, "doublingTimeMonths": 5.7, "startYear": 2010, "endYear": 2026, "n": 466 }, "frontierByYear": [ { "year": 1950, "maxLog10Compute": 1.602, "computeFlop": 40, "model": "Theseus", "organization": "Bell Laboratories" }, { "year": 1957, "maxLog10Compute": 5.842, "computeFlop": 694894.9377361819, "model": "Perceptron Mark I", "organization": "Cornell Aeronautical Laboratory,Cornell University" }, { "year": 1959, "maxLog10Compute": 8.778, "computeFlop": 600000000, "model": "Pandemonium (morse)", "organization": "Massachusetts Institute of Technology (MIT)" }, { "year": 1960, "maxLog10Compute": 8.857, "computeFlop": 720000000, "model": "Perceptron (1960)", "organization": "Cornell Aeronautical Laboratory" }, { "year": 1962, "maxLog10Compute": 6.193, "computeFlop": 1559250, "model": "Linear Decision Functions", "organization": "Bell Laboratories" }, { "year": 1963, "maxLog10Compute": 7.352, "computeFlop": 22500000, "model": "Print Recognition Logic", "organization": "IBM" }, { "year": 1965, "maxLog10Compute": 6.033, "computeFlop": 1080000, "model": "Heuristic Reinforcement Learning", "organization": "Purdue University" }, { "year": 1966, "maxLog10Compute": 8.025, "computeFlop": 105917060, "model": "LTE speaker verification system", "organization": "IBM" }, { "year": 1975, "maxLog10Compute": 6.715, "computeFlop": 5184000, "model": "Cognitron", "organization": "Biological Cybernetics" }, { "year": 1980, "maxLog10Compute": 8.437, "computeFlop": 273738240, "model": "Neocognitron", "organization": "NHK Broadcasting Science Research Laboratories" }, { "year": 1983, "maxLog10Compute": 8.511, "computeFlop": 324000000, "model": "ASE+ACE", "organization": "University of Massachusetts Amherst" }, { "year": 1986, "maxLog10Compute": 8.829, "computeFlop": 673920000, "model": "MLP with back-propagation", "organization": "University of California San Diego,Carnegie Mellon University (CMU)" }, { "year": 1987, "maxLog10Compute": 10.452, "computeFlop": 28328002560, "model": "NetTalk (transcription)", "organization": "Princeton University" }, { "year": 1988, "maxLog10Compute": 8.472, "computeFlop": 296425000, "model": "MLN-ASR", "organization": "McGill University" }, { "year": 1989, "maxLog10Compute": 12.175, "computeFlop": 1496338054440, "model": "Zip CNN", "organization": "AT&T,Bell Laboratories" }, { "year": 1990, "maxLog10Compute": 10.896, "computeFlop": 78775200000, "model": "SexNet compression", "organization": "Unknown" }, { "year": 1991, "maxLog10Compute": 10.878, "computeFlop": 75474000000, "model": "Weight Decay", "organization": "Unknown" }, { "year": 1992, "maxLog10Compute": 13.261, "computeFlop": 18232157622832.703, "model": "TD-Gammon", "organization": "IBM" }, { "year": 1993, "maxLog10Compute": 13.11, "computeFlop": 12869570138112, "model": "Siamese-TDNN", "organization": "Bell Laboratories" }, { "year": 1994, "maxLog10Compute": 13.27, "computeFlop": 18621900000000, "model": "Predictive Coding NN", "organization": "Technical University of Munich" }, { "year": 1995, "maxLog10Compute": 11.291, "computeFlop": 195544800000, "model": "LISSOM", "organization": "University of Texas at Austin" }, { "year": 1996, "maxLog10Compute": 11.945, "computeFlop": 881733600000, "model": "MUSIC perceptron", "organization": "Unknown" }, { "year": 1997, "maxLog10Compute": 13.498, "computeFlop": 31512000000000, "model": "LSTM", "organization": "Technical University of Munich" }, { "year": 1998, "maxLog10Compute": 12.449, "computeFlop": 2810937600000, "model": "LeNet-5", "organization": "AT&T" }, { "year": 1999, "maxLog10Compute": 12.904, "computeFlop": 8013600000000, "model": "RECONTRA-categorized", "organization": "Unknown" }, { "year": 2000, "maxLog10Compute": 15.802, "computeFlop": 6339000000000000, "model": "Neural LM", "organization": "University of Montreal / Université de Montréal" }, { "year": 2001, "maxLog10Compute": 13.799, "computeFlop": 63000000000000, "model": "Decision tree (classification)", "organization": "Mitsubishi Electric Research Labs,Compaq CRL" }, { "year": 2003, "maxLog10Compute": 15.222, "computeFlop": 1666869200000000, "model": "NPLM (AP News)", "organization": "University of Montreal / Université de Montréal" }, { "year": 2004, "maxLog10Compute": 15.444, "computeFlop": 2782080000000000, "model": "LMICA", "organization": "Unknown" }, { "year": 2005, "maxLog10Compute": 14.064, "computeFlop": 115848000000000, "model": "Hierarchical LM", "organization": "Unknown" }, { "year": 2006, "maxLog10Compute": 14.872, "computeFlop": 745200000000000, "model": "SVM-CNN", "organization": "New York University (NYU)" }, { "year": 2007, "maxLog10Compute": 18.161, "computeFlop": 1449446400000000000, "model": "SB-LM", "organization": "Google" }, { "year": 2008, "maxLog10Compute": 9.208, "computeFlop": 1614600000, "model": "GNN", "organization": "University of Siena" }, { "year": 2009, "maxLog10Compute": 15.621, "computeFlop": 4181852160000000, "model": "LCNP MNIST", "organization": "Unknown" }, { "year": 2010, "maxLog10Compute": 16.732, "computeFlop": 53960000000000000, "model": "RNN LM", "organization": "Johns Hopkins University" }, { "year": 2011, "maxLog10Compute": 16.716, "computeFlop": 52000000000000000, "model": "CNN Committee (MNIST)", "organization": "IDSIA" }, { "year": 2012, "maxLog10Compute": 17.778, "computeFlop": 600000000000000000, "model": "Unsupervised High-level Feature Learner", "organization": "Google" }, { "year": 2013, "maxLog10Compute": 18.417, "computeFlop": 2612736000000000000, "model": "DistBelief NNLM", "organization": "Google" }, { "year": 2014, "maxLog10Compute": 20.474, "computeFlop": 297600000001000000000, "model": "SNM-skip", "organization": "Google" }, { "year": 2015, "maxLog10Compute": 20.58, "computeFlop": 380000000000000000000, "model": "AlphaGo Fan", "organization": "DeepMind" }, { "year": 2016, "maxLog10Compute": 21.821, "computeFlop": 6.620000000001e+21, "model": "GNMT", "organization": "Google" }, { "year": 2017, "maxLog10Compute": 20.926, "computeFlop": 843000000000000000000, "model": "JFT", "organization": "Google Research,Carnegie Mellon University (CMU)" }, { "year": 2018, "maxLog10Compute": 21.942, "computeFlop": 8.74395e+21, "model": "ResNeXt-101 32x48d", "organization": "Facebook" }, { "year": 2019, "maxLog10Compute": 23.032, "computeFlop": 1.0773400001e+23, "model": "AlphaStar", "organization": "DeepMind" }, { "year": 2020, "maxLog10Compute": 23.497, "computeFlop": 3.14e+23, "model": "GPT-3 175B (davinci)", "organization": "OpenAI" }, { "year": 2021, "maxLog10Compute": 24.311, "computeFlop": 2.047e+24, "model": "FLAN 137B", "organization": "Google Research" }, { "year": 2022, "maxLog10Compute": 24.438, "computeFlop": 2.7415e+24, "model": "Minerva (540B)", "organization": "Google" }, { "year": 2023, "maxLog10Compute": 25.699, "computeFlop": 5.0000000001e+25, "model": "Gemini 1.0 Ultra", "organization": "Google DeepMind" }, { "year": 2024, "maxLog10Compute": 25.58, "computeFlop": 3.8e+25, "model": "Llama 3.1-405B", "organization": "Meta AI" }, { "year": 2025, "maxLog10Compute": 26.699, "computeFlop": 5.0000000000001e+26, "model": "Grok 4", "organization": "xAI" }, { "year": 2026, "maxLog10Compute": 25.588, "computeFlop": 3.87e+25, "model": "Composer 2.5", "organization": "Cursor" } ] }, "costTrend": [ { "year": 2012, "n": 4, "medianCostUsd2023": 11.761154415598087, "maxCostUsd2023": 16.029257461780738, "topModel": "Unsupervised High-level Feature Learner" }, { "year": 2013, "n": 3, "medianCostUsd2023": 30.028097313317172, "maxCostUsd2023": 3255.1839637787602, "topModel": "DistBelief NNLM" }, { "year": 2014, "n": 2, "medianCostUsd2023": 145.44487886566205, "maxCostUsd2023": 239.2376655010467, "topModel": "VGG16" }, { "year": 2015, "n": 3, "medianCostUsd2023": 1394.115708804708, "maxCostUsd2023": 4828.322276803044, "topModel": "AlphaGo Fan" }, { "year": 2016, "n": 6, "medianCostUsd2023": 17146.646723076294, "maxCostUsd2023": 201331.569937687, "topModel": "GNMT" }, { "year": 2017, "n": 6, "medianCostUsd2023": 10892.441078921935, "maxCostUsd2023": 471445.3247973037, "topModel": "AlphaGo Master" }, { "year": 2018, "n": 8, "medianCostUsd2023": 1343.4035298450158, "maxCostUsd2023": 134077.42688552555, "topModel": "ResNeXt-101 32x48d" }, { "year": 2019, "n": 16, "medianCostUsd2023": 61735.524695856555, "maxCostUsd2023": 4328311.470477086, "topModel": "OpenAI Five" }, { "year": 2020, "n": 19, "medianCostUsd2023": 5021.241075362514, "maxCostUsd2023": 2116866.2428390156, "topModel": "GPT-3 175B (davinci)" }, { "year": 2021, "n": 34, "medianCostUsd2023": 36609.25978364835, "maxCostUsd2023": 3848505.6256471733, "topModel": "Megatron-Turing NLG 530B" }, { "year": 2022, "n": 28, "medianCostUsd2023": 94645.7963359519, "maxCostUsd2023": 4897551.634095244, "topModel": "GPT-3.5 (davinci-002)" }, { "year": 2023, "n": 34, "medianCostUsd2023": 163238.51801259865, "maxCostUsd2023": 37334304.976625234, "topModel": "GPT-4 (Mar 2023)" }, { "year": 2024, "n": 8, "medianCostUsd2023": 17690564.946113274, "maxCostUsd2023": 52885433.95402246, "topModel": "Llama 3.1-405B" }, { "year": 2025, "n": 7, "medianCostUsd2023": 44588963.624007165, "maxCostUsd2023": 387842678.08636147, "topModel": "Grok 4" } ], "powerTrend": [ { "year": 2011, "n": 1, "medianPowerW": 246.22598104391827, "maxPowerW": 246.22598104391827 }, { "year": 2012, "n": 1, "medianPowerW": 2116.2977934480996, "maxPowerW": 2116.2977934480996 }, { "year": 2014, "n": 2, "medianPowerW": 1213.849268210018, "maxPowerW": 2137.653074766455 }, { "year": 2015, "n": 3, "medianPowerW": 285.60949180561556, "maxPowerW": 290.9065505304121 }, { "year": 2016, "n": 5, "medianPowerW": 4200.468649711526, "maxPowerW": 37828.64014138488 }, { "year": 2017, "n": 5, "medianPowerW": 4180.032869488383, "maxPowerW": 32873.78910003616 }, { "year": 2018, "n": 11, "medianPowerW": 4964.696746005496, "maxPowerW": 209159.05867416784 }, { "year": 2019, "n": 15, "medianPowerW": 262131.94609302413, "maxPowerW": 944347.7271739373 }, { "year": 2020, "n": 19, "medianPowerW": 39011.038545758856, "maxPowerW": 5100759.605963441 }, { "year": 2021, "n": 32, "medianPowerW": 181104.9158497605, "maxPowerW": 3714163.0079064844 }, { "year": 2022, "n": 32, "medianPowerW": 218549.75620916547, "maxPowerW": 4198431.396906301 }, { "year": 2023, "n": 37, "medianPowerW": 102249.42137571001, "maxPowerW": 38423900.11232233 }, { "year": 2024, "n": 18, "medianPowerW": 189354.88660292566, "maxPowerW": 22622532.159299146 }, { "year": 2025, "n": 8, "medianPowerW": 2105605.9601618517, "maxPowerW": 109948656.20196915 } ], "orgLeaderboard": [ { "organization": "OpenAI", "orgCategory": "Industry", "country": "United States", "modelCount": 65, "computeKnownCount": 22, "frontierCount": 7, "recentCount": 44, "openWeightsCount": 8, "maxComputeFlop": 3.8e+26, "latestDate": "2026-04-23", "medianLog10Compute": 22.846 }, { "organization": "Google", "orgCategory": "Industry", "country": "United States", "modelCount": 54, "computeKnownCount": 27, "frontierCount": 16, "recentCount": 2, "openWeightsCount": 7, "maxComputeFlop": 7.34e+24, "latestDate": "2025-08-26", "medianLog10Compute": 21.912 }, { "organization": "Google DeepMind", "orgCategory": "Industry", "country": "United States", "modelCount": 39, "computeKnownCount": 6, "frontierCount": 1, "recentCount": 31, "openWeightsCount": 4, "maxComputeFlop": 5.0000000001e+25, "latestDate": "2026-04-15", "medianLog10Compute": 22.153 }, { "organization": "Alibaba", "orgCategory": "Industry", "country": "China", "modelCount": 32, "computeKnownCount": 16, "frontierCount": 0, "recentCount": 29, "openWeightsCount": 22, "maxComputeFlop": 1.512e+25, "latestDate": "2026-05-19", "medianLog10Compute": 24.513 }, { "organization": "DeepMind", "orgCategory": "Industry", "country": "United Kingdom", "modelCount": 30, "computeKnownCount": 22, "frontierCount": 5, "recentCount": 0, "openWeightsCount": 1, "maxComputeFlop": 6.31e+23, "latestDate": "2022-12-01", "medianLog10Compute": 21.066 }, { "organization": "Meta AI", "orgCategory": "Industry", "country": "United States", "modelCount": 29, "computeKnownCount": 19, "frontierCount": 2, "recentCount": 13, "openWeightsCount": 21, "maxComputeFlop": 5.18400000000001e+25, "latestDate": "2026-04-08", "medianLog10Compute": 23.74 }, { "organization": "Microsoft", "orgCategory": "Industry", "country": "United States", "modelCount": 17, "computeKnownCount": 8, "frontierCount": 0, "recentCount": 1, "openWeightsCount": 4, "maxComputeFlop": 4.831e+22, "latestDate": "2026-06-02", "medianLog10Compute": 21.021 }, { "organization": "Anthropic", "orgCategory": "Industry", "country": "United States", "modelCount": 16, "computeKnownCount": 3, "frontierCount": 2, "recentCount": 15, "openWeightsCount": 0, "maxComputeFlop": 3.35e+25, "latestDate": "2026-06-09", "medianLog10Compute": 25.431 }, { "organization": "Stanford University", "orgCategory": "Academia", "country": "United States", "modelCount": 16, "computeKnownCount": 4, "frontierCount": 1, "recentCount": 0, "openWeightsCount": 1, "maxComputeFlop": 78328627000000000000, "latestDate": "2021-10-31", "medianLog10Compute": 15.577 }, { "organization": "NVIDIA", "orgCategory": "Industry", "country": "United States", "modelCount": 15, "computeKnownCount": 12, "frontierCount": 3, "recentCount": 9, "openWeightsCount": 6, "maxComputeFlop": 1.8e+25, "latestDate": "2026-06-04", "medianLog10Compute": 23.468 }, { "organization": "University of Toronto", "orgCategory": "Academia", "country": "Canada", "modelCount": 15, "computeKnownCount": 5, "frontierCount": 3, "recentCount": 0, "openWeightsCount": 0, "maxComputeFlop": 470000000000000000, "latestDate": "2014-11-10", "medianLog10Compute": 16.565 }, { "organization": "Google Research", "orgCategory": "Industry", "country": "United States", "modelCount": 13, "computeKnownCount": 8, "frontierCount": 2, "recentCount": 2, "openWeightsCount": 2, "maxComputeFlop": 2.5272e+24, "latestDate": "2023-12-23", "medianLog10Compute": 23.285 }, { "organization": "Baidu", "orgCategory": "Industry", "country": "China", "modelCount": 12, "computeKnownCount": 4, "frontierCount": 0, "recentCount": 4, "openWeightsCount": 5, "maxComputeFlop": 2.25e+22, "latestDate": "2025-03-16", "medianLog10Compute": 21.149 }, { "organization": "Google Brain", "orgCategory": "Industry", "country": "United States", "modelCount": 12, "computeKnownCount": 6, "frontierCount": 2, "recentCount": 0, "openWeightsCount": 3, "maxComputeFlop": 1.12e+23, "latestDate": "2022-05-23", "medianLog10Compute": 20.776 }, { "organization": "Facebook AI Research", "orgCategory": "Industry", "country": "United States, France", "modelCount": 11, "computeKnownCount": 7, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 5, "maxComputeFlop": 3.366e+22, "latestDate": "2021-11-11", "medianLog10Compute": 20.663 }, { "organization": "Carnegie Mellon University (CMU)", "orgCategory": "Academia", "country": "United States", "modelCount": 10, "computeKnownCount": 10, "frontierCount": 2, "recentCount": 1, "openWeightsCount": 3, "maxComputeFlop": 1.1e+21, "latestDate": "2023-08-28", "medianLog10Compute": 17.138 }, { "organization": "University of Oxford", "orgCategory": "Academia", "country": "United Kingdom", "modelCount": 10, "computeKnownCount": 4, "frontierCount": 2, "recentCount": 0, "openWeightsCount": 3, "maxComputeFlop": 29000000000000000000, "latestDate": "2022-12-19", "medianLog10Compute": 19.066 }, { "organization": "AT&T", "orgCategory": "Industry", "country": "United States", "modelCount": 9, "computeKnownCount": 2, "frontierCount": 2, "recentCount": 0, "openWeightsCount": 0, "maxComputeFlop": 2810937600000, "latestDate": "2009-09-21", "medianLog10Compute": 11.854 }, { "organization": "ByteDance", "orgCategory": "Industry", "country": "China", "modelCount": 9, "computeKnownCount": 2, "frontierCount": 0, "recentCount": 9, "openWeightsCount": 1, "maxComputeFlop": 2.505e+25, "latestDate": "2026-02-12", "medianLog10Compute": 24.771 }, { "organization": "DeepSeek", "orgCategory": "Industry", "country": "China", "modelCount": 9, "computeKnownCount": 8, "frontierCount": 0, "recentCount": 9, "openWeightsCount": 9, "maxComputeFlop": 9.702e+24, "latestDate": "2026-04-24", "medianLog10Compute": 24.519 } ], "countryLeaderboard": [ { "country": "United States", "countryShort": "United States", "iso3": "USA", "modelCount": 664, "computeKnownCount": 328, "frontierCount": 77, "recentCount": 189, "openWeightsCount": 198, "maxComputeFlop": 5.0000000000001e+26, "orgCount": 278 }, { "country": "China", "countryShort": "China", "iso3": "CHN", "modelCount": 178, "computeKnownCount": 97, "frontierCount": 4, "recentCount": 104, "openWeightsCount": 91, "maxComputeFlop": 2.505e+25, "orgCount": 91 }, { "country": "South Korea", "countryShort": "South Korea", "iso3": "KOR", "modelCount": 24, "computeKnownCount": 15, "frontierCount": 1, "recentCount": 12, "openWeightsCount": 14, "maxComputeFlop": 3.85e+24, "orgCount": 16 }, { "country": "France", "countryShort": "France", "iso3": "FRA", "modelCount": 50, "computeKnownCount": 30, "frontierCount": 1, "recentCount": 8, "openWeightsCount": 27, "maxComputeFlop": 2.13e+25, "orgCount": 31 }, { "country": "United Kingdom", "countryShort": "United Kingdom", "iso3": "GBR", "modelCount": 89, "computeKnownCount": 46, "frontierCount": 9, "recentCount": 6, "openWeightsCount": 16, "maxComputeFlop": 6.31e+23, "orgCount": 39 }, { "country": "Singapore", "countryShort": "Singapore", "iso3": null, "modelCount": 16, "computeKnownCount": 8, "frontierCount": 1, "recentCount": 5, "openWeightsCount": 9, "maxComputeFlop": 3.036551985824e+24, "orgCount": 14 }, { "country": "Hong Kong", "countryShort": "Hong Kong", "iso3": null, "modelCount": 15, "computeKnownCount": 9, "frontierCount": 0, "recentCount": 5, "openWeightsCount": 6, "maxComputeFlop": 3.036551985824e+24, "orgCount": 12 }, { "country": "Canada", "countryShort": "Canada", "iso3": "CAN", "modelCount": 57, "computeKnownCount": 27, "frontierCount": 7, "recentCount": 4, "openWeightsCount": 8, "maxComputeFlop": 4.3e+23, "orgCount": 35 }, { "country": "Israel", "countryShort": "Israel", "iso3": "ISR", "modelCount": 12, "computeKnownCount": 8, "frontierCount": 1, "recentCount": 2, "openWeightsCount": 7, "maxComputeFlop": 3.7e+23, "orgCount": 11 }, { "country": "Japan", "countryShort": "Japan", "iso3": "JPN", "modelCount": 11, "computeKnownCount": 7, "frontierCount": 1, "recentCount": 2, "openWeightsCount": 5, "maxComputeFlop": 2.5562e+23, "orgCount": 10 }, { "country": "Saudi Arabia", "countryShort": "Saudi Arabia", "iso3": "SAU", "modelCount": 4, "computeKnownCount": 2, "frontierCount": 0, "recentCount": 2, "openWeightsCount": 0, "maxComputeFlop": 1.05e+25, "orgCount": 4 }, { "country": "UAE", "countryShort": "UAE", "iso3": "ARE", "modelCount": 3, "computeKnownCount": 3, "frontierCount": 1, "recentCount": 2, "openWeightsCount": 3, "maxComputeFlop": 3.76e+24, "orgCount": 2 }, { "country": "Germany", "countryShort": "Germany", "iso3": "DEU", "modelCount": 31, "computeKnownCount": 20, "frontierCount": 4, "recentCount": 1, "openWeightsCount": 14, "maxComputeFlop": 8.46e+22, "orgCount": 26 }, { "country": "Switzerland", "countryShort": "Switzerland", "iso3": "CHE", "modelCount": 12, "computeKnownCount": 6, "frontierCount": 2, "recentCount": 1, "openWeightsCount": 2, "maxComputeFlop": 478000000000000000, "orgCount": 7 }, { "country": "Australia", "countryShort": "Australia", "iso3": "AUS", "modelCount": 6, "computeKnownCount": 3, "frontierCount": 0, "recentCount": 1, "openWeightsCount": 2, "maxComputeFlop": 8.46e+22, "orgCount": 6 }, { "country": "Finland", "countryShort": "Finland", "iso3": "FIN", "modelCount": 6, "computeKnownCount": 2, "frontierCount": 0, "recentCount": 1, "openWeightsCount": 4, "maxComputeFlop": 2.42784e+21, "orgCount": 4 }, { "country": "Netherlands", "countryShort": "Netherlands", "iso3": "NLD", "modelCount": 5, "computeKnownCount": 4, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 2, "maxComputeFlop": 8.46e+22, "orgCount": 5 }, { "country": "Multinational", "countryShort": "Multinational", "iso3": null, "modelCount": 4, "computeKnownCount": 0, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 1, "maxComputeFlop": 0, "orgCount": 2 }, { "country": "Czechia", "countryShort": "Czechia", "iso3": "CZE", "modelCount": 3, "computeKnownCount": 1, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 0, "maxComputeFlop": 14463360000000000000, "orgCount": 3 }, { "country": "Italy", "countryShort": "Italy", "iso3": "ITA", "modelCount": 3, "computeKnownCount": 2, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 1, "maxComputeFlop": 4531200000, "orgCount": 3 }, { "country": "Belgium", "countryShort": "Belgium", "iso3": "BEL", "modelCount": 2, "computeKnownCount": 1, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 0, "maxComputeFlop": 982800000000000, "orgCount": 2 }, { "country": "Denmark", "countryShort": "Denmark", "iso3": "DNK", "modelCount": 2, "computeKnownCount": 1, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 0, "maxComputeFlop": 578102400000000000, "orgCount": 2 }, { "country": "India", "countryShort": "India", "iso3": "IND", "modelCount": 2, "computeKnownCount": 2, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 1, "maxComputeFlop": 8.46e+22, "orgCount": 2 }, { "country": "Poland", "countryShort": "Poland", "iso3": "POL", "modelCount": 2, "computeKnownCount": 2, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 0, "maxComputeFlop": 93939056640000000000, "orgCount": 2 }, { "country": "Spain", "countryShort": "Spain", "iso3": "ESP", "modelCount": 2, "computeKnownCount": 2, "frontierCount": 1, "recentCount": 0, "openWeightsCount": 1, "maxComputeFlop": 8.46e+22, "orgCount": 2 }, { "country": "Sweden", "countryShort": "Sweden", "iso3": "SWE", "modelCount": 2, "computeKnownCount": 1, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 1, "maxComputeFlop": 8.46e+22, "orgCount": 2 }, { "country": "Taiwan", "countryShort": "Taiwan", "iso3": "TWN", "modelCount": 2, "computeKnownCount": 2, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 0, "maxComputeFlop": 85800000000000000000, "orgCount": 2 }, { "country": "Argentina", "countryShort": "Argentina", "iso3": "ARG", "modelCount": 1, "computeKnownCount": 0, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 0, "maxComputeFlop": 0, "orgCount": 1 }, { "country": "Austria", "countryShort": "Austria", "iso3": "AUT", "modelCount": 1, "computeKnownCount": 0, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 1, "maxComputeFlop": 0, "orgCount": 1 }, { "country": "Croatia", "countryShort": "Croatia", "iso3": "HRV", "modelCount": 1, "computeKnownCount": 0, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 0, "maxComputeFlop": 0, "orgCount": 1 }, { "country": "Iran", "countryShort": "Iran", "iso3": "IRN", "modelCount": 1, "computeKnownCount": 0, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 0, "maxComputeFlop": 0, "orgCount": 1 }, { "country": "Ireland", "countryShort": "Ireland", "iso3": "IRL", "modelCount": 1, "computeKnownCount": 1, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 0, "maxComputeFlop": 34153451000000000, "orgCount": 1 }, { "country": "Malaysia", "countryShort": "Malaysia", "iso3": "MYS", "modelCount": 1, "computeKnownCount": 0, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 0, "maxComputeFlop": 0, "orgCount": 1 }, { "country": "Norway", "countryShort": "Norway", "iso3": "NOR", "modelCount": 1, "computeKnownCount": 0, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 1, "maxComputeFlop": 0, "orgCount": 1 }, { "country": "Russia", "countryShort": "Russia", "iso3": "RUS", "modelCount": 1, "computeKnownCount": 1, "frontierCount": 0, "recentCount": 0, "openWeightsCount": 0, "maxComputeFlop": 51880000000000000, "orgCount": 1 } ], "accessibilityMix": { "openWeights": 219, "closed": 211, "unknown": 98 }, "fullCatalogAccessibilityMix": { "openWeights": 318, "closed": 452, "unknown": 260 }, "domainMix": [ { "domain": "Language", "count": 299 }, { "domain": "Vision", "count": 147 }, { "domain": "Multimodal", "count": 41 }, { "domain": "Biology", "count": 34 }, { "domain": "Games", "count": 27 }, { "domain": "Speech", "count": 22 }, { "domain": "Image generation", "count": 21 }, { "domain": "Video", "count": 13 }, { "domain": "Robotics", "count": 10 }, { "domain": "Earth science", "count": 7 }, { "domain": "Other", "count": 7 }, { "domain": "Audio", "count": 5 }, { "domain": "Mathematics", "count": 4 }, { "domain": "Medicine", "count": 2 }, { "domain": "Search", "count": 2 }, { "domain": "Materials science", "count": 1 }, { "domain": "Recommendation", "count": 1 } ] }, "caveats": [ "Coverage is not exhaustive; Epoch curates 'notable' models.", "Compute estimates carry uncertainty (see Confidence field).", "Future-dated entries reflect the source snapshot and are shown as-is.", "Descriptive trends, not predictions.", "Multi-country collaboration models are co-attributed to each participating country in the country leaderboard.", "Epoch AI's 'Frontier model' flag reflects top-10 training compute at release time, not capability or impact.", "Org/country model counts use the full dated catalog. Compute-derived metrics (frontierCount, maxComputeFlop) use only the compute-known subset.", "openWeightsCount is a proxy for tracked open-release activity only." ], "meta": { "generatedAt": "2026-07-18T00:55:13.689Z", "asOf": "2026-07-18", "source": { "name": "Epoch AI — Notable AI Models", "publisher": "Epoch AI", "url": "https://epoch.ai/data/ai-models" }, "version": "1.0.0" } }