| Benchmark | GPT-4o | GPT-5.6 Sol |
|---|---|---|
| AA Agentic Index | 9.7 | 57.8 |
| AA Intelligence | 11.2 | 61 |
| AA-LCR | 55 | 77.7 |
| AA-Omniscience | -10.5 | 22 |
| arc_agi_1 | 4.5 | 97.5 |
| ARC-AGI-2 | 0 | 92.5 |
| Artificial Analysis Coding Index | 24.2 | 78.3 |
| critpt | 0 | 32.3 |
| gdpval | 0 | 61.1 |
| GPQA Diamond | 70.1 | 94.6 |
| HLE | 5.3 | 49.5 |
| IFBench | 36 | 72.7 |
| longbench_v2 | 48.1 | 67.1 |
| mathvision | 30.4 | 95.8 |
| MMMU-Pro | 59.9 | 84.6 |
| OmniScience Accuracy | 23.7 | 59.4 |
| OmniScience Non-Hallucination | 62.1 | 10.6 |
| OSWorld-Verified | 5 | 83 |
| scicode | 33.4 | 56.9 |
| simplebench | 17.8 | 64.8 |
| Terminal-Bench Hard | 8.3 | 65.9 |
| Video-MME | 77.2 | 89.5 |
| τ²-Bench Telecom (AA run) | 28.9 | 85.1 |
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.