| GPQA Diamond | 88.1% |
| HLE(text-only subset only) | 42% |
| Terminal-Bench 2.1 | 88.2% |
| Terminal-Bench 3.0 | 28.3% |
| Terminal-Bench 4.0 | 37.9% |
| DeepSWE v1.1 | 66.9% |
| ProgramBench | 19% |
| NL2Repo-Bench | 58% |
| CyberGym | 84.5% |
| ExploitGym | 15% |
| HLE (w/tools) | 62.5% |
| Automation-Bench | 48.8% |
| Agents' Last Exam | 28.5% |