| Benchmark | GPT-5.1 | o3 |
|---|---|---|
| AA Agentic Index | 32.2 | 36.1 |
| AA Intelligence | 37.5 | 31.1 |
| AA-LCR | 76.7 | 73.3 |
| AA-Omniscience | 5.4 | -15.3 |
| AIME 2025 | 94.2 | 89.2 |
| arc_agi_1 | 72.8 | 60.8 |
| ARC-AGI-2 | 17.6 | 6.5 |
| arena_vision | 1250 | 1217 |
| Artificial Analysis Coding Index | 49.4 | 38.4 |
| browsecomp | 50.8 | 49.7 |
| critpt | 4.9 | 1.1 |
| Fortress | 25.7 | 16 |
| frontiermath_tier_4 | 12.5 | 2.1 |
| gdpval | 25 | 12.8 |
| GPQA Diamond | 88.1 | 83.3 |
| HLE | 28.5 | 20.3 |
| hmmt_feb_2025 | 96.3 | 77.5 |
| IFBench | 72.9 | 71.4 |
| MMLU-Pro | 87 | 85 |
| MMMU | 85.4 | 82.9 |
| MMMU-Pro | 75.5 | 76.4 |
| multichallenge | 63.4 | 60.4 |
| MultiNRC | 49 | 45.5 |
| OmniScience Accuracy | 37.7 | 38.6 |
| OmniScience Non-Hallucination | 48.1 | 12.9 |
| scicode | 43.3 | 41 |
| simplebench | 53.2 | 53.1 |
| simpleqa_verified | 48.9 | 53 |
| swe_bench_bash | 66 | 58.4 |
| SWE-bench Verified | 76.3 | 69.1 |
| Tau2 airline | 67 | 64.8 |
| Tau2 retail | 77.9 | 80.2 |
| Terminal-Bench Hard | 45.5 | 37.1 |
| τ²-Bench Telecom (AA run) | 81.9 | 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.