| Benchmark | o4-mini | Opus 5 |
|---|---|---|
| AA Agentic Index | 36.1 | 59.2 |
| AA Intelligence | 26.1 | 63.1 |
| AA-LCR | 60 | 78.7 |
| AA-Omniscience | -35.7 | 37.1 |
| arc_agi_1 | 58.7 | 97.5 |
| ARC-AGI-2 | 6.1 | 90.4 |
| Artificial Analysis Coding Index | 25.6 | 78 |
| browsecomp | 51.5 | 90.8 |
| critpt | 0.6 | 29.1 |
| gdpval | 25.4 | 67.2 |
| GPQA Diamond | 81.4 | 93.7 |
| HLE | 16.5 | 64.7 |
| MMMU-Pro | 69.2 | 84.7 |
| OmniScience Accuracy | 24.8 | 60.9 |
| OmniScience Non-Hallucination | 19.5 | 40.5 |
| scicode | 46.5 | 55.7 |
| simplebench | 38.7 | 80.6 |
| simpleqa_verified | 23.9 | 56.7 |
| SWE-bench Verified | 68.1 | 96 |
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.