| GPQA Diamond | 92.4% |
| HLE(text-only subset only) | 42.7% |
| Codeforces (Rating) | 3348 |
| MathArena Apex | 65.3% |
| Terminal-Bench 2.1 | 87.9% |
| Terminal-Bench 3.0 | 11.8% |
| Terminal-Bench 4.0 | 12.4% |
| DeepSWE v1.1 | 62.7% |
| ProgramBench | 15.5% |
| NL2Repo-Bench | 61.5% |
| CyberGym | 83.3% |
| SEC-Bench Pro | 56.4% |
| ExploitGym | 5.4% |
| HLE (w/tools) | 60% |
| Automation-Bench | 43.2% |
| Agents' Last Exam | 25.7% |