| Benchmark | Claude Opus 4.5 | GPT-5.6 Sol |
|---|---|---|
| AA Agentic Index | 59.6 | 57.8 |
| AA Intelligence | 41.9 | 61 |
| AA-LCR | 76 | 77.7 |
| AA-Omniscience | 14 | 22 |
| aime_2026 | 95.1 | 99.9 |
| arc_agi_1 | 82 | 97.5 |
| ARC-AGI-2 | 37.6 | 92.5 |
| arena_elo | 1473 | 1482 |
| arena_text_factuality | 1458 | 1474 |
| Artificial Analysis Coding Index | 47.8 | 78.3 |
| browsecomp | 37 | 90.4 |
| charxiv_rq | 68.5 | 87.8 |
| coding_arena_elo | 1495 | 1619 |
| critpt | 4.6 | 32.3 |
| cybergym | 50.6 | 83.6 |
| gdpval | 47.3 | 61.1 |
| GPQA Diamond | 87 | 94.6 |
| HLE | 30.8 | 49.5 |
| HLE (with tools) | 43.4 | 58 |
| IFBench | 58 | 72.7 |
| livebench_agentic_coding | 39.7 | 56.2 |
| livebench_coding | 79.7 | 83.9 |
| livebench_data_analysis | 74.4 | 79.8 |
| livebench_instruction_following | 62.5 | 71.8 |
| livebench_language | 81.3 | 87.7 |
| livebench_math | 90.4 | 96.2 |
| livebench_reasoning | 80.1 | 91.7 |
| longbench_v2 | 64.4 | 67.1 |
| mathvision | 77.1 | 95.8 |
| MCP Atlas | 69.8 | 83.6 |
| MMMU-Pro | 74 | 84.6 |
| MMVU | 77.3 | 81.2 |
| OmniScience Accuracy | 46.6 | 59.4 |
| OmniScience Non-Hallucination | 39 | 10.6 |
| OSWorld-Verified | 66.3 | 83 |
| PaperBench | 72.9 | 90.5 |
| scicode | 50 | 56.9 |
| simplebench | 62 | 64.8 |
| simpleqa_verified | 41.8 | 71.6 |
| SWE-bench Pro | 57.1 | 64.6 |
| Terminal-Bench Hard | 47 | 65.9 |
| toolathlon | 43.5 | 58 |
| τ²-Bench Telecom (AA run) | 89.5 | 85.1 |
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.