| Benchmark | Claude Opus 4.6 | GPT-5.6 Sol |
|---|---|---|
| AA Agentic Index | 67.6 | 57.8 |
| AA Intelligence | 44.9 | 61 |
| AA-LCR | 74.3 | 77.7 |
| AA-Omniscience | 13.7 | 22 |
| aime_2026 | 96.7 | 99.9 |
| arc_agi_1 | 93 | 97.5 |
| arc_agi_3 | 0.5 | 7.8 |
| ARC-AGI-2 | 68.8 | 92.5 |
| arena_elo | 1497 | 1482 |
| arena_text_factuality | 1487 | 1474 |
| Artificial Analysis Coding Index | 48.1 | 78.3 |
| baby_vision_with_python | 38.4 | 88.9 |
| browsecomp | 86.8 | 90.4 |
| charxiv_rq | 69.1 | 87.8 |
| coding_arena_elo | 1545 | 1619 |
| critpt | 12.6 | 32.3 |
| cybergym | 73.8 | 83.6 |
| gdpval | 55.9 | 61.1 |
| GPQA Diamond | 91.3 | 94.6 |
| HiL-Bench | 38.3 | 32.3 |
| HLE | 53.1 | 49.5 |
| HLE (with tools) | 62.7 | 58 |
| IFBench | 62.5 | 72.7 |
| livebench_agentic_coding | 49 | 56.2 |
| livebench_coding | 78.2 | 83.9 |
| livebench_data_analysis | 69.9 | 79.8 |
| livebench_instruction_following | 63.3 | 71.8 |
| livebench_language | 83.3 | 87.7 |
| livebench_math | 89.3 | 96.2 |
| livebench_reasoning | 88.7 | 91.7 |
| mathvision | 71.2 | 95.8 |
| MCP Atlas | 76.8 | 83.6 |
| MMMU-Pro | 77.3 | 84.6 |
| officeqa_pro | 57.1 | 63.2 |
| OmniScience Accuracy | 47 | 59.4 |
| OmniScience Non-Hallucination | 37.2 | 10.6 |
| OSWorld-Verified | 72.7 | 83 |
| scicode | 52 | 56.9 |
| simplebench | 67.6 | 64.8 |
| simpleqa_verified | 46.5 | 71.6 |
| SWE-bench Pro | 57.3 | 64.6 |
| Terminal-Bench Hard | 65.4 | 65.9 |
| toolathlon | 56.8 | 58 |
| τ²-Bench Telecom (AA run) | 99.3 | 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.