Native computer use, 1M context, first to surpass human baseline on OSWorld
| Benchmark | Score | Rank |
|---|---|---|
OSWorld Computer use in real desktop environments | 75% | #1 / 6 |
FrontierMath Unpublished research-level mathematics problems | 47.6% | #1 / 4 |
Terminal Agentic terminal coding tasks requiring multi-step execution | 75.1% | #2 / 37 |
ARC-AGI Novel reasoning tasks requiring fluid intelligence | 73.3% | #3 / 21 |
GPQA PhD-level science questions even experts struggle with | 92.8% | #4 / 54 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 84% | #11 / 31 |
MMMUvals.ai College-level multimodal reasoning across 30+ disciplines | 79.2% | #14 / 33 |
MMLU-Provals.ai Harder 10-option successor to MMLU; more reasoning-focused | 84.6% | #17 / 30 |