1M context with parallel agent teams, #1 on Arena Elo at 1504
| Benchmark | Score | Rank |
|---|---|---|
SWE-bench Real-world GitHub issue resolution | 80.8% | #3 / 39 |
OSWorld Computer use in real desktop environments | 72.7% | #5 / 13 |
ARC-AGI Novel reasoning tasks requiring fluid intelligence | 68.8% | #8 / 26 |
Arena Elo Human preference ranking via blind comparisons | 1504 | #9 / 51 |
HumanEval Coding ability - generating correct Python functions | 95% | #10 / 50 |
MMLU-Provals.ai Harder 10-option successor to MMLU; more reasoning-focused | 89.1% | #10 / 45 |
Terminal Agentic terminal coding tasks requiring multi-step execution | 65.4% | #12 / 56 |
MMMUvals.ai College-level multimodal reasoning across 30+ disciplines | 83.9% | #14 / 46 |
GPQA PhD-level science questions even experts struggle with | 91.3% | #15 / 73 |
MMLU Tests knowledge across 57 subjects from STEM to humanities | 91% | #17 / 54 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 84.7% | #19 / 48 |
hleArtificial Analysis | 36.7% | #19 / 61 |
MATH Competition-level mathematics problems | 93% | #22 / 50 |