| Benchmark | Claude Opus 4.7 | GPT-4o mini |
|---|---|---|
| AA Agentic Index | 64.6 | 1 |
| AA Intelligence | 55 | 7 |
| ARC-AGI-2 | 75.8 | 0 |
| Artificial Analysis Coding Index | 73.6 | 11.4 |
| gaia_overall | 28 | 4.7 |
| gdpval | 58.6 | 0 |
| GPQA Diamond | 94.2 | 42.6 |
| harmbench | 17.5 | 84.9 |
| HLE | 54.7 | 4.2 |
| IFBench | 58.6 | 31 |
| MMMU-Pro | 78.8 | 41.5 |
| scicode | 54.5 | 22.9 |
| simplebench | 61.7 | 10.7 |
| SWE-bench Verified | 87.6 | 8.7 |
| TauBench V3 - Banking | 34.6 | 2.9 |
| Terminal-Bench 2.1 | 83.1 | 5.6 |
| τ³-Bench | 28.9 | 2.9 |
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.