New frontier leader. SWE-bench Pro 69.2% tops GPT-5.5 and Gemini 3.1 Pro; ~4x less likely to let code flaws pass unremarked. Same pricing as 4.7, fast mode 2.5x speed.
| Benchmark | Score | Rank |
|---|---|---|
SWE-bench Real-world GitHub issue resolution | 88.6% | #1 / 40 |
OSWorld Computer use in real desktop environments | 83.4% | #1 / 9 |
Arena Elo Human preference ranking via blind comparisons | 1890 | #1 / 52 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 87.8% | #3 / 40 |
Terminal Agentic terminal coding tasks requiring multi-step execution | 74.6% | #4 / 48 |
MMLU-Provals.ai Harder 10-option successor to MMLU; more reasoning-focused | 89.6% | #4 / 38 |
MMMUvals.ai College-level multimodal reasoning across 30+ disciplines | 86.6% | #7 / 39 |
GPQA PhD-level science questions even experts struggle with | 92% | #9 / 64 |