| agents-on-automationbench-public | Beam | – | Success Rate: 37.0 |
| code-generation-on-deepswe | Beam | – | Pass@1: 0.444 |
| code-generation-on-scicode | Beam | – | Accuracy: 49.7 |
| code-generation-on-swe-bench-multilingual | Beam | – | Accuracy: 78.0 |
| code-generation-on-swe-bench-pro | Beam | – | Resolved: 65.5 |
| code-generation-on-swe-bench-verified | Beam | – | Accuracy: 80.9 |
| code-generation-on-terminal-bench-2-1 | Beam | – | Accuracy: 80.1 |
| deep-research-agents-on-browsecomp | Beam (w/ context management) | – | Accuracy: 77.4 |
| instruction-following-on-ifbench | Beam | – | Accuracy: 79.7 |
| long-context-understanding-on-longbench-v2 | Beam | – | Accuracy: 65.5 |
| mathematical-reasoning-on-aime-2026 | Beam | – | Accuracy: 97.8 |
| multiple-choice-question-answering-mcqa-on-gpqa-diamond | Beam | – | Accuracy: 90.5 |
| physics-question-answering-on-critpt | Beam | – | Accuracy: 16.3 |
| question-answering-on-humanity-s-last-exam-hle | Beam | – | Accuracy: 36.2 |