| Benchmark | Claude Opus 4.6 | Opus 5 |
|---|---|---|
| AA Agentic Index | 67.6 | 59.2 |
| AA Intelligence | 44.9 | 63.1 |
| AA-LCR | 74.3 | 78.7 |
| AA-Omniscience | 13.7 | 37.1 |
| arc_agi_1 | 93 | 97.5 |
| arc_agi_3 | 0.5 | 30.2 |
| ARC-AGI-2 | 68.8 | 90.4 |
| arena_elo | 1497 | 1487 |
| arena_text_factuality | 1487 | 1481 |
| Artificial Analysis Coding Index | 48.1 | 78 |
| browsecomp | 86.8 | 90.8 |
| coding_arena_elo | 1545 | 1662 |
| critpt | 12.6 | 29.1 |
| gdpval | 55.9 | 67.2 |
| GDPval-AA (Elo) | 1619 | 1827 |
| GPQA Diamond | 91.3 | 93.7 |
| HiL-Bench | 38.3 | 57 |
| HLE | 53.1 | 64.7 |
| HLE (with tools) | 62.7 | 64.7 |
| Legal Agent Benchmark | 4.2 | 23 |
| livebench_agentic_coding | 49 | 65.2 |
| livebench_coding | 78.2 | 81.5 |
| livebench_data_analysis | 69.9 | 74.5 |
| livebench_instruction_following | 63.3 | 63.8 |
| livebench_language | 83.3 | 88.7 |
| livebench_math | 89.3 | 95.7 |
| livebench_reasoning | 88.7 | 91.2 |
| MCP Atlas | 76.8 | 85.8 |
| MMMU-Pro | 77.3 | 84.7 |
| officeqa | 73.5 | 78.1 |
| officeqa_pro | 57.1 | 66.9 |
| OmniScience Accuracy | 47 | 60.9 |
| OmniScience Non-Hallucination | 37.2 | 40.5 |
| OSWorld-Verified | 72.7 | 4 |
| scicode | 52 | 55.7 |
| simplebench | 67.6 | 80.6 |
| simpleqa_verified | 46.5 | 56.7 |
| swe_bench_multimodal | 27.1 | 59.4 |
| SWE-bench Multilingual | 77.8 | 89.5 |
| SWE-bench Pro | 57.3 | 79.2 |
| SWE-bench Verified | 80.8 | 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.