| agents-on-toolathlon-verified-108-tasks | Claude Opus 5 | #1 | Pass@1: 0.806 |
| code-generation-on-swe-bench-verified | Claude Opus 5 | #1 | Accuracy: 96.0 |
| computer-use-agents-on-osworld-2-0 | Claude Opus 5 (Anthropic system card) | #1 | Success Rate: 70.57 |
| multi-task-language-understanding-on-include | Claude Opus 5 | #1 | Accuracy: 89.8 |
| code-generation-on-swe-bench-pro | Claude Opus 5 | #3 | Resolved: 79.2 |
| exploit-generation-on-exploitbench | Claude Opus 5 (ExploitBench AutoNudge) | #3 | Coverage: 0.7 |
| code-generation-on-swe-bench-multilingual | Claude Opus 5 | #4 | Accuracy: 89.5 |
| medical-question-answering-on-healthbench-professional | Claude Opus 5 | #4 | Score: 59.8 |
| abstract-reasoning-on-arc-agi-2 | Claude Opus 5 | #5 | Accuracy: 90.42 |
| abstract-reasoning-on-arc-agi | Claude Opus 5 | #7 | Accuracy: 97.5 |
| code-generation-on-deepswe | Claude Opus 5 | #11 | Pass@1: 0.688 |