Training compute growth
Maximum disclosed training compute per year on a log10(FLOP) scale. Missing values remain absent; the current year is incomplete.
Training compute growth
Total 468 · Page 18/19 · Observed 9/25/2026, 3:30:45 AM
| Name | Group | Reported value | Date | Notes | Source |
|---|---|---|---|---|---|
| TC-DNN-BLSTM-DNN | Carnegie Mellon University (CMU) | 194,101,919,999,999,970 FLOP | 2015-04-06 | Unreleased | Source ↗ |
| MMLSTM (WT-2) | Beijing University of Posts and Telecommunications,University of West London | 193,800,000,000,000,000 FLOP | 2019-12-05 | Unreleased | Source ↗ |
| Mogrifier RLSTM (WT2) | DeepMind | 140,000,000,000,000,000 FLOP | 2022-11-03 | Unreleased | Source ↗ |
| Fold2Seq | IBM,Texas A&M | 140,000,000,000,000,000 FLOP | 2021-06-24 | Unreleased | Source ↗ |
| EnhanceNet | Max Planck Institute for Intelligent Systems | 130,792,319,999,999,980 FLOP | 2016-12-23 | Open weights (unrestricted) | Source ↗ |
| ReLU-Speech | Google,University of Toronto,New York University (NYU) | 127,733,760,000,000,000 FLOP | 2013-05-26 | — | Source ↗ |
| Dropout-LSTM+Noise(Bernoulli) (WT2) | Columbia University,New York University (NYU),Princeton University | 127,000,000,000,000,000 FLOP | 2018-05-03 | Unreleased | Source ↗ |
| Fractional Max-Pooling | University of Warwick | 100,000,000,000,000,000 FLOP | 2014-12-18 | — | Source ↗ |
| Named Entity Recognition model | Carnegie Mellon University (CMU) | 96,940,800,000,000,000 FLOP | 2016-03-04 | — | Source ↗ |
| aLSTM(depth-2)+RecurrentPolicy (WT2) | University of Manchester,Alan Turing Institute | 72,960,000,000,000,000 FLOP | 2018-05-22 | Unreleased | Source ↗ |
| SmooCT | University College London (UCL) | 69,000,000,000,000,000 FLOP | 2014-07-01 | — | Source ↗ |
| Pluribus | Facebook AI Research | 66,000,000,000,000,000 FLOP | 2019-07-11 | Unreleased | Source ↗ |
| MMLSTM (PTB) | Beijing University of Posts and Telecommunications,University of West London | 58,298,782,000,000,000 FLOP | 2019-12-05 | Unreleased | Source ↗ |
| Swift | Intel Labs | 53,370,000,000,000,000 FLOP | 2023-08-30 | Unreleased | Source ↗ |
| Fine-tuned-AWD-LSTM-DOC (fin) | Samsung R&D Institute Russia | 51,880,000,000,000,000 FLOP | 2018-11-12 | Unreleased | Source ↗ |
| U-Net | University of Freiburg | 50,832,252,000,000,000 FLOP | 2015-05-18 | Open weights (unrestricted) | Source ↗ |
| PeptideBERT | Carnegie Mellon University (CMU) | 49,000,000,000,000,000 FLOP | 2023-08-28 | Open weights (unrestricted) | Source ↗ |
| SPN-4+KN5 | Singapore University of Technology & Design,DSO National Laboratories | 44,000,000,000,000,000 FLOP | 2014-09-14 | Unreleased | Source ↗ |
| StyleGAN | Nvidia | 39,300,000,000,000,000 FLOP | 2018-12-12 | Open weights (non-commercial) | Source ↗ |
| Word2Vec (large) | 38,880,000,000,000,000 FLOP | 2013-10-16 | — | Source ↗ | |
| genCNN + dyn eval | Chinese Academy of Sciences,Huawei Noah's Ark Lab,Dublin City University | 34,153,451,000,000,000 FLOP | 2015-03-17 | Unreleased | Source ↗ |
| VD-LSTM+REAL Large | Salesforce Research,Stanford University | 21,300,000,000,000,000 FLOP | 2016-11-04 | Unreleased | Source ↗ |
| SPIDER2 | Griffith University,University of Iowa,Dezhou University | 18,220,000,000,000,000 FLOP | 2016-10-28 | Open weights (non-commercial) | Source ↗ |
| LSTM LM | RWTH Aachen University | 16,600,000,000,000,000 FLOP | 2012-09-09 | — | Source ↗ |
| RNTN | Stanford University | 14,220,000,000,000,000 FLOP | 2013-10-01 | Unreleased | Source ↗ |
Data: Epoch AI · CC BY · View source · Automatically synced; not an exhaustive global sample