| Benchmark | Claude Sonnet 5 | Gemini 2.5 Flash |
|---|---|---|
| AA Agentic Index | 49.7 | 18.8 |
| AA Intelligence | 55.3 | 17.7 |
| AA-LCR | 77 | 65.7 |
| AA-Omniscience | 16.4 | -29.8 |
| arena_text_factuality | 1453 | 1437 |
| Artificial Analysis Coding Index | 71.5 | 22.2 |
| critpt | 16.9 | 1.4 |
| gdpval | 54.8 | 11.9 |
| GPQA Diamond | 91.1 | 82.8 |
| HLE | 57.4 | 12.1 |
| MMMU-Pro | 77.3 | 69.1 |
| OmniScience Accuracy | 40 | 26.1 |
| OmniScience Non-Hallucination | 60.6 | 24.7 |
| scicode | 53.6 | 35.9 |
| simplebench | 60.6 | 41.2 |
| SWE-bench Verified | 85.2 | 60.4 |
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.