| Benchmark | GPT-5.6 Sol | o3 |
|---|---|---|
| AA Agentic Index | 57.8 | 36.1 |
| AA Intelligence | 61 | 31.1 |
| AA-LCR | 77.7 | 73.3 |
| AA-Omniscience | 22 | -15.3 |
| arc_agi_1 | 97.5 | 60.8 |
| ARC-AGI-2 | 92.5 | 6.5 |
| Artificial Analysis Coding Index | 78.3 | 38.4 |
| browsecomp | 90.4 | 49.7 |
| critpt | 32.3 | 1.1 |
| gdpval | 61.1 | 12.8 |
| GPQA Diamond | 94.6 | 83.3 |
| HLE | 49.5 | 20.3 |
| IFBench | 72.7 | 71.4 |
| longbench_v2 | 67.1 | 58.8 |
| MMMU-Pro | 84.6 | 76.4 |
| OmniScience Accuracy | 59.4 | 38.6 |
| OmniScience Non-Hallucination | 10.6 | 12.9 |
| scicode | 56.9 | 41 |
| simplebench | 64.8 | 53.1 |
| simpleqa_verified | 71.6 | 53 |
| strongreject | 98.5 | 97 |
| Terminal-Bench Hard | 65.9 | 37.1 |
| τ²-Bench Telecom (AA run) | 85.1 | 80.7 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.