| Benchmark | Claude Sonnet 5 | GPT-5.6 Sol |
|---|---|---|
| AA Agentic Index | 49.7 | 57.8 |
| AA Intelligence | 55.3 | 61 |
| AA-LCR | 77 | 77.7 |
| AA-Omniscience | 16.4 | 22 |
| arena_elo | 1463 | 1482 |
| arena_text_factuality | 1453 | 1474 |
| Artificial Analysis Coding Index | 71.5 | 78.3 |
| browsecomp | 84.7 | 90.4 |
| charxiv_rq | 77 | 87.8 |
| coding_arena_elo | 1540 | 1619 |
| critpt | 16.9 | 32.3 |
| DeepSWE 1.1 | 54 | 73 |
| GDP (Surge AI) | 81.6 | 30.7 |
| gdpval | 54.8 | 61.1 |
| GDPval-AA v2 | 1618 | 1748 |
| GPQA Diamond | 91.1 | 94.6 |
| healthbench_professional | 57.8 | 60.5 |
| HLE | 57.4 | 49.5 |
| livebench_agentic_coding | 59.4 | 56.2 |
| livebench_coding | 80.7 | 83.9 |
| livebench_data_analysis | 71.7 | 79.8 |
| livebench_instruction_following | 63.9 | 71.8 |
| livebench_language | 75 | 87.7 |
| livebench_math | 92.9 | 96.2 |
| livebench_reasoning | 88.7 | 91.7 |
| MMMU-Pro | 77.3 | 84.6 |
| officeqa_pro | 59.4 | 63.2 |
| OmniScience Accuracy | 40 | 59.4 |
| OmniScience Non-Hallucination | 60.6 | 10.6 |
| OSWorld-Verified | 81.2 | 83 |
| scicode | 53.6 | 56.9 |
| simplebench | 60.6 | 64.8 |
| simpleqa_verified | 25 | 71.6 |
| SWE-bench Pro | 63.2 | 64.6 |
| TauBench V3 - Banking | 37.3 | 44.3 |
| Terminal-Bench 2.1 | 80.5 | 89.5 |
| Terminal-bench 3.0 | 14.6 | 34.6 |
| toolathlon | 54.3 | 58 |
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.