Balanced GPT-5.6 model for everyday professional work with strong capability and lower cost than Sol.
| Benchmark | Score | Rank |
|---|---|---|
Terminal Agentic terminal coding tasks requiring multi-step execution | 87.4% | #2 / 56 |
AA Coding Independent composite of implementation, terminal use, and real-codebase agent performance | 77.4% | #3 / 18 |
ARC-AGI Novel reasoning tasks requiring fluid intelligence | 91.3% | #3 / 30 |
OSWorld Computer use in real desktop environments | 50.2% | #11 / 13 |
AA Intelligence Independent composite of agentic work, coding, scientific reasoning, and general capability | 55% | #12 / 18 |
MMMU College-level multimodal reasoning across 30+ disciplines | 86.5% | #16 / 57 |
hle | 40% | #19 / 76 |
LiveCodeBench Contamination-free competitive programming (filtered by cutoff date) | 85.9% | #20 / 59 |
GPQA PhD-level science questions even experts struggle with | 90.8% | #27 / 88 |
MMLU-Pro Harder 10-option successor to MMLU; more reasoning-focused | 86.7% | #34 / 58 |