| Benchmark | GPT-4o mini | GPT-5.1 |
|---|---|---|
| AA Agentic Index | 1 | 32.2 |
| AA Intelligence | 7 | 37.5 |
| ARC-AGI-2 | 0 | 17.6 |
| Artificial Analysis Coding Index | 11.4 | 49.4 |
| Fortress | 48.1 | 25.7 |
| gdpval | 0 | 25 |
| GPQA Diamond | 42.6 | 88.1 |
| HLE | 4.2 | 28.5 |
| IFBench | 31 | 72.9 |
| MMMU | 59.4 | 85.4 |
| MMMU-Pro | 41.5 | 75.5 |
| scicode | 22.9 | 43.3 |
| simplebench | 10.7 | 53.2 |
| SWE-bench Verified | 8.7 | 76.3 |
| TauBench V3 - Banking | 2.9 | 15.9 |
| Terminal-Bench 2.1 | 5.6 | 52.4 |
| τ³-Bench | 2.9 | 14 |
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.