Deep reasoning model, strongest on abstract math at launch
| Benchmark | Score | Rank |
|---|---|---|
HumanEval Coding ability - generating correct Python functions | 95.2% | #9 / 50 |
HellaSwag Common sense reasoning about everyday situations | 96.4% | #9 / 36 |
MMLU Tests knowledge across 57 subjects from STEM to humanities | 92.4% | #10 / 54 |
ARC-C Grade-school science questions requiring reasoning | 98.1% | #10 / 40 |
ARC-AGIARC Prize Novel reasoning tasks requiring fluid intelligence | 21.2% | #17 / 26 |
MATH Competition-level mathematics problems | 94% | #20 / 50 |
SWE-bench Real-world GitHub issue resolution | 68.4% | #26 / 39 |
Arena Elo Human preference ranking via blind comparisons | 1408 | #28 / 51 |
GPQA PhD-level science questions even experts struggle with | 88% | #31 / 73 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 79% | #34 / 48 |
TerminalArtificial Analysis Agentic terminal coding tasks requiring multi-step execution | 37.9% | #35 / 56 |
hleArtificial Analysis | 23.9% | #36 / 61 |
MMMUvals.ai College-level multimodal reasoning across 30+ disciplines | 72.8% | #37 / 46 |
MMLU-Provals.ai Harder 10-option successor to MMLU; more reasoning-focused | 79.7% | #40 / 45 |