Improved agentic reliability with better tool use and planning
| Benchmark | Score | Rank |
|---|---|---|
ARC-C Grade-school science questions requiring reasoning | 97.6% | #13 / 40 |
MMLU-Provals.ai Harder 10-option successor to MMLU; more reasoning-focused | 87.9% | #13 / 45 |
HumanEval Coding ability - generating correct Python functions | 94.5% | #14 / 50 |
SWE-bench Real-world GitHub issue resolution | 74.5% | #18 / 39 |
HellaSwag Common sense reasoning about everyday situations | 92.8% | #19 / 36 |
MMLU Tests knowledge across 57 subjects from STEM to humanities | 89.5% | #22 / 54 |
MATH Competition-level mathematics problems | 88.4% | #26 / 50 |
MMMUvals.ai College-level multimodal reasoning across 30+ disciplines | 77.5% | #28 / 46 |
Arena Elo Human preference ranking via blind comparisons | 1372 | #31 / 51 |
TerminalArtificial Analysis Agentic terminal coding tasks requiring multi-step execution | 34.3% | #40 / 56 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 66.5% | #42 / 48 |
GPQA PhD-level science questions even experts struggle with | 80.9% | #48 / 73 |
hleArtificial Analysis | 11.9% | #48 / 61 |