Best coding model of 2024, dominated SWE-bench for months
| Benchmark | Score | Rank |
|---|---|---|
HumanEval Coding ability - generating correct Python functions | 93.7% | #17 / 50 |
ARC-C Grade-school science questions requiring reasoning | 96.7% | #23 / 40 |
HellaSwag Common sense reasoning about everyday situations | 89% | #27 / 36 |
MMLU Tests knowledge across 57 subjects from STEM to humanities | 88.7% | #28 / 54 |
MATH Competition-level mathematics problems | 78.3% | #33 / 50 |
MMMUvals.ai College-level multimodal reasoning across 30+ disciplines | 68.8% | #33 / 39 |
MMLU-Provals.ai Harder 10-option successor to MMLU; more reasoning-focused | 78.4% | #36 / 38 |
SWE-bench Real-world GitHub issue resolution | 49% | #37 / 40 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 49.6% | #37 / 40 |
Arena Elo Human preference ranking via blind comparisons | 1280 | #44 / 52 |
GPQA PhD-level science questions even experts struggle with | 65% | #50 / 64 |