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 8/19 · Observed 9/25/2026, 3:30:45 AM
| Name | Group | Reported value | Date | Notes | Source |
|---|---|---|---|---|---|
| Meta Pseudo Labels | Google Brain,Google AI | 47,900,000,000,000,000,000,000 FLOP | 2021-03-01 | Unreleased | Source ↗ |
| GShard (dense) | 47,650,000,000,000,000,000,000 FLOP | 2020-06-30 | Unreleased | Source ↗ | |
| Eagle 2 | Nvidia,Nanjing University,Tsinghua University,Hong Kong Polytechnic University,Johns Hopkins University,New York University (NYU) | 47,156,000,000,000,000,000,000 FLOP | 2025-01-20 | Open weights (non-commercial) | — |
| DALL-E | OpenAI | 47,000,000,000,000,000,000,000 FLOP | 2021-01-05 | API access | Source ↗ |
| Volcano 13B | Korea University,Korea Advanced Institute of Science and Technology (KAIST),LG | 45,600,000,000,000,000,000,000 FLOP | 2023-11-13 | Open weights (non-commercial) | Source ↗ |
| CoAtNet | Google,Google Research,Google Brain | 42,700,000,000,000,000,000,000 FLOP | 2021-06-09 | Unreleased | Source ↗ |
| AlphaFold 3 | Google DeepMind,Isomorphic Labs | 41,405,645,000,000,000,000,000 FLOP | 2024-05-08 | Open weights (non-commercial) | Source ↗ |
| BASIC-L | 41,200,000,000,000,000,000,000 FLOP | 2021-11-19 | Unreleased | Source ↗ | |
| RDT-1B | Tsinghua University | 40,600,000,000,000,000,000,000 FLOP | 2024-10-10 | Open weights (unrestricted) | Source ↗ |
| Pangu-Weather | Huawei | 39,800,000,000,000,000,000,000 FLOP | 2023-07-05 | Open weights (non-commercial) | Source ↗ |
| ProtBERT-BFD | Technical University of Munich,Nvidia,Seoul National University,Google,Oak Ridge National Laboratory,Med AI Technology | 39,000,000,000,000,000,000,000 FLOP | 2021-05-04 | Open weights (unrestricted) | Source ↗ |
| Student of Games | DeepMind | 36,679,273,004,682,866,000,000 FLOP | 2021-12-06 | Unreleased | Source ↗ |
| Qwen3-Omni-30B-A3B | Alibaba | 36,000,000,000,000,000,000,000 FLOP | 2025-09-22 | Open weights (unrestricted) | Source ↗ |
| PLUG | Alibaba | 35,997,696,000,000,000,000,000 FLOP | 2021-04-19 | Hosted access (no API) | Source ↗ |
| GQA-8-XXL | Google Research | 34,912,896,000,000,000,000,000 FLOP | 2023-12-23 | Unreleased | Source ↗ |
| XLMR-XXL | Facebook AI Research | 33,660,000,000,000,000,000,000 FLOP | 2021-08-17 | Open weights (unrestricted) | Source ↗ |
| T5-11B | 33,000,000,000,000,000,000,000 FLOP | 2019-10-23 | Open weights (unrestricted) | Source ↗ | |
| SPHINX (Llama 2 13B) | Shanghai AI Lab,Chinese University of Hong Kong (CUHK),ShanghaiTech University | 30,400,000,000,000,000,000,000 FLOP | 2023-11-13 | Open weights (restricted use) | Source ↗ |
| Turing ULRv5 | Microsoft | 28,983,951,000,000,000,000,000 FLOP | 2021-09-28 | Hosted access (no API) | Source ↗ |
| CogView | Tsinghua University,Alibaba DAMO Academy | 26,800,000,000,000,000,000,000 FLOP | 2021-05-26 | Open weights (unrestricted) | Source ↗ |
| Noisy Student (L2) | Carnegie Mellon University (CMU),Google | 26,120,000,000,000,000,000,000 FLOP | 2019-11-11 | Unreleased | Source ↗ |
| ALIGN | Google Research | 25,986,700,000,010,000,000,000 FLOP | 2021-06-11 | Unreleased | Source ↗ |
| DeBERTa | Microsoft | 25,880,000,000,000,000,000,000 FLOP | 2021-06-10 | Open weights (unrestricted) | Source ↗ |
| GOAT | DeepMind | 24,120,000,000,000,000,000,000 FLOP | 2021-07-27 | Unreleased | Source ↗ |
| XGLM-7.5B | Meta AI,Facebook AI Research | 22,500,000,000,000,000,000,000 FLOP | 2021-12-20 | Open weights (non-commercial) | Source ↗ |
Data: Epoch AI · CC BY · View source · Automatically synced; not an exhaustive global sample