| Benchmark | Claude Sonnet 4.6 | GPT-4o mini |
|---|---|---|
| AA Agentic Index | 61.6 | 1 |
| AA Intelligence | 48.4 | 7 |
| ARC-AGI-2 | 60.4 | 0 |
| Artificial Analysis Coding Index | 63 | 11.4 |
| gdpval | 54.8 | 0 |
| GPQA Diamond | 89.9 | 42.6 |
| harmbench | 24.2 | 84.9 |
| HLE | 49 | 4.2 |
| IFBench | 56.6 | 31 |
| MMMU-Pro | 75.6 | 41.5 |
| scicode | 47 | 22.9 |
| SWE-bench Verified | 79.6 | 8.7 |
| TauBench V3 - Banking | 34.4 | 2.9 |
| Terminal-Bench 2.1 | 71.2 | 5.6 |
| τ³-Bench | 30.5 | 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.