Full reasoning model with tool use, first to break 80% on GPQA Diamond
| Benchmark | Score | Rank |
|---|---|---|
MATH Competition-level mathematics problems | 91.6% | #23 / 50 |
SWE-bench Real-world GitHub issue resolution | 71.7% | #25 / 39 |
ARC-AGIARC Prize Novel reasoning tasks requiring fluid intelligence | 4.9% | #25 / 26 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 71.5% | #39 / 48 |
MMLU-Provals.ai Harder 10-option successor to MMLU; more reasoning-focused | 78.7% | #41 / 45 |
GPQA PhD-level science questions even experts struggle with | 83.3% | #44 / 73 |