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 17/19 · Observed 9/25/2026, 3:30:45 AM
| Name | Group | Reported value | Date | Notes | Source |
|---|---|---|---|---|---|
| AlphaX-1 | Facebook AI Research,Brown University | 889,344,000,000,000,000 FLOP | 2019-10-02 | Unreleased | Source ↗ |
| WeNet (Penn Treebank) | Amazon | 730,000,001,000,000,000 FLOP | 2019-04-08 | Unreleased | Source ↗ |
| R-FCN | Tsinghua University,Microsoft Research | 719,357,760,000,000,000 FLOP | 2016-06-21 | — | Source ↗ |
| QRNN | Salesforce Research | 688,664,720,000,000,000 FLOP | 2018-02-01 | Unreleased | Source ↗ |
| (ensemble): AWD-LSTM-DOC (fin) × 5 (WT2) | NTT Communication Science Laboratories,Tohoku University | 666,000,000,000,000,000 FLOP | 2018-08-30 | Open weights (unrestricted) | Source ↗ |
| UDSMProt | Fraunhofer Heinrich Hertz Institute | 637,000,000,000,000,000 FLOP | 2019-09-04 | Open weights (unrestricted) | Source ↗ |
| Unsupervised High-level Feature Learner | 600,000,000,000,000,000 FLOP | 2012-07-12 | — | Source ↗ | |
| 4 layer QRNN (h=2500) | Salesforce Research | 591,588,150,000,000,000 FLOP | 2018-03-22 | Unreleased | Source ↗ |
| DeepLoc | Technical University of Denmark,University of Copenhagen | 578,102,400,000,000,000 FLOP | 2017-07-07 | Hosted access (no API) | Source ↗ |
| CT-MoS (WT2) | Google,National Tsing Hua University | 540,000,000,000,000,000 FLOP | 2020-12-25 | Unreleased | Source ↗ |
| Visualizing CNNs | New York University (NYU) | 532,000,000,000,000,000 FLOP | 2013-11-12 | — | Source ↗ |
| GANs | University of Montreal / Université de Montréal | 518,400,000,000,000,000 FLOP | 2014-06-10 | — | Source ↗ |
| DCNN | University of Maryland,Rutgers University | 480,986,910,000,000,000 FLOP | 2015-08-07 | Unreleased | Source ↗ |
| DNN EM segmentation | IDSIA,SUPSI | 478,000,000,000,000,000 FLOP | 2012-12-03 | Unreleased | Source ↗ |
| AlexNet | University of Toronto | 470,000,000,000,000,000 FLOP | 2012-09-30 | — | Source ↗ |
| GL-LWGC-AWD-MoS-LSTM + dynamic evaluation (WT2) | Ben-Gurion University of the Negev | 456,000,000,000,000,000 FLOP | 2017-08-29 | Unreleased | Source ↗ |
| Big-Little Net (speech) | IBM | 429,004,800,000,000,000 FLOP | 2018-07-10 | Open weights (unrestricted) | Source ↗ |
| AWD-LSTM-DRILL + dynamic evaluation† (WT2) | IDIAP | 408,000,000,000,000,000 FLOP | 2019-05-14 | Open weights (unrestricted) | Source ↗ |
| AWD-LSTM + MoS + Partial Shuffled | University of Texas at Austin | 315,000,000,000,000,000 FLOP | 2019-06-10 | Open weights (non-commercial) | Source ↗ |
| DistBelief Speech | 311,400,000,000,000,000 FLOP | 2012-12-03 | — | Source ↗ | |
| Fraternal dropout + AWD-LSTM 3-layer (WT2) | Jagiellonian University,Mila - Quebec AI (originally Montreal Institute for Learning Algorithms),University of Montreal / Université de Montréal | 306,000,000,000,000,000 FLOP | 2017-10-31 | Unreleased | Source ↗ |
| AWD-LSTM - 3-layer LSTM (tied) + continuous cache pointer (WT2) | Salesforce Research | 297,000,000,000,000,000 FLOP | 2017-08-07 | Unreleased | Source ↗ |
| Dropout (ImageNet) | University of Toronto | 273,196,800,000,000,000 FLOP | 2012-06-03 | Unreleased | Source ↗ |
| Big-Little Net | IBM | 246,048,000,000,000,000 FLOP | 2018-07-10 | Open weights (unrestricted) | Source ↗ |
| Hierarchical Scene Labeling (Stanford Background) | New York University (NYU) | 237,746,880,000,000,000 FLOP | 2013-08-01 | Unreleased | Source ↗ |
Data: Epoch AI · CC BY · View source · Automatically synced; not an exhaustive global sample