| Benchmark | GPT-5.2 | o1 |
|---|---|---|
| AA Agentic Index | 60.2 | 31.1 |
| AA Intelligence | 43.3 | 24 |
| AA-LCR | 79.3 | 63.3 |
| AA-Omniscience | 0.3 | -10.6 |
| AIME 2025 | 100 | 81.7 |
| arena_vision | 1245 | 1193 |
| Artificial Analysis Coding Index | 48.7 | 39.7 |
| critpt | 11.6 | 0.3 |
| Fortress | 17.5 | 19.4 |
| gdpval | 48.3 | 11.5 |
| GPQA Diamond | 92.4 | 75.7 |
| HLE | 45.5 | 7.7 |
| IFBench | 75.4 | 70.3 |
| livebench | 74.8 | 52.3 |
| MathVista | 82.8 | 71.8 |
| metr_hcast | 77 | 42.6 |
| metr_re_bench | 35 | 0 |
| metr_swaa | 99.6 | 100 |
| mmmlu | 89.6 | 87.7 |
| OmniScience Accuracy | 44.3 | 34.7 |
| OmniScience Non-Hallucination | 38.4 | 30.7 |
| scicode | 52.1 | 35.8 |
| SEAL VISTA | 46.6 | 45.3 |
| simplebench | 45.8 | 41.7 |
| SWE-bench Verified | 80 | 48.9 |
| Terminal-Bench Hard | 62.2 | 12.9 |
| τ²-Bench Telecom (AA run) | 98.7 | 62.6 |
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.