Unified reasoning and chat, 400K context, 94.6% on AIME 2025
| Benchmark | Score | Rank |
|---|---|---|
HellaSwag Common sense reasoning about everyday situations | 97.2% | #4 / 36 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 86.6% | #4 / 31 |
GPQA PhD-level science questions even experts struggle with | 88.4% | #9 / 54 |
ARC-C Grade-school science questions requiring reasoning | 98.1% | #9 / 40 |
HumanEval Coding ability - generating correct Python functions | 94.8% | #11 / 49 |
ARC-AGIARC Prize Novel reasoning tasks requiring fluid intelligence | 18.3% | #13 / 21 |
MMLU Tests knowledge across 57 subjects from STEM to humanities | 91.2% | #14 / 53 |
SWE-bench Real-world GitHub issue resolution | 74.9% | #15 / 38 |
MMMUvals.ai College-level multimodal reasoning across 30+ disciplines | 78.9% | #15 / 33 |
MATH Competition-level mathematics problems | 94.6% | #17 / 49 |
TerminalArtificial Analysis Agentic terminal coding tasks requiring multi-step execution | 37.9% | #17 / 37 |
Arena Elo Human preference ranking via blind comparisons | 1398 | #21 / 41 |
MMLU-Provals.ai Harder 10-option successor to MMLU; more reasoning-focused | 82.2% | #23 / 30 |