| Benchmark | Model | Rank | Results |
|---|---|---|---|
| code-generation-on-humaneval | Atlas (cluster) | – | Pass@1: 91.5 |
| code-generation-on-mbpp | Atlas (cluster) | – | Accuracy: 83.6 |
| mathematical-reasoning-on-aime-2025 | Atlas (RL) (pool extension) | – | Accuracy: 40.0 |
| question-answering-on-natural-questions | Atlas (RL) (pool extension) | – | EM: 45.3 |