| GPQA Diamond | 92.9% |
| HLE | 43.5% |
| MathArena Apex | 65.6% |
| Terminal-Bench 2.1 | 88.3% |
| Terminal-Bench 3.0 | 17.7% |
| Terminal-Bench 4.0 | 12.6% |
| DeepSWE v1.1 | 67.5% |
| ProgramBench | 17.5% |
| NL2Repo-Bench | 58% |
| CyberGym | 80% |
| HLE (w/tools) | 59.8% |
| Automation-Bench | 46.7% |
| Agents' Last Exam | 27.6% |
| Chartography (w/tools) | 68.1% |
| BabyVision (w/tools) | 85.7% |
| ZeroBench-main (w/tools) | 41% |