| GPQA Diamond | 94.1% |
| HLE | 44.5% |
| Terminal-Bench 2.1 | 88.8% |
| Terminal-Bench 3.0 | 34.4% |
| Terminal-Bench 4.0 | 39.9% |
| DeepSWE v1.1 | 73% |
| ProgramBench | 23% |
| NL2Repo-Bench | 56.8% |
| CyberGym | 84.5% |
| SEC-Bench Pro | 74.3% |
| ExploitGym | 33.7% |
| Automation-Bench | 45.8% |
| Agents' Last Exam | 26.7% |
| Chartography (w/tools) | 79.9% |
| BabyVision (w/tools) | 88.9% |
| ZeroBench-main (w/tools) | 53% |