| Benchmark | o3 | Opus 5 |
|---|---|---|
| AA Agentic Index | 36.1 | 59.2 |
| AA Intelligence | 31.1 | 63.1 |
| AA-LCR | 73.3 | 78.7 |
| AA-Omniscience | -15.3 | 37.1 |
| arc_agi_1 | 60.8 | 97.5 |
| ARC-AGI-2 | 6.5 | 90.4 |
| Artificial Analysis Coding Index | 38.4 | 78 |
| browsecomp | 49.7 | 90.8 |
| critpt | 1.1 | 29.1 |
| gdpval | 12.8 | 67.2 |
| GPQA Diamond | 83.3 | 93.7 |
| HLE | 20.3 | 64.7 |
| MMMU-Pro | 76.4 | 84.7 |
| OmniScience Accuracy | 38.6 | 60.9 |
| OmniScience Non-Hallucination | 12.9 | 40.5 |
| scicode | 41 | 55.7 |
| simplebench | 53.1 | 80.6 |
| simpleqa_verified | 53 | 56.7 |
| SWE-bench Verified | 69.1 | 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.