| Benchmark | Claude Opus 4.5 | Opus 5 |
|---|---|---|
| AA Agentic Index | 59.6 | 59.2 |
| AA Intelligence | 41.9 | 63.1 |
| AA-LCR | 76 | 78.7 |
| AA-Omniscience | 14 | 37.1 |
| arc_agi_1 | 82 | 97.5 |
| ARC-AGI-2 | 37.6 | 90.4 |
| arena_elo | 1473 | 1487 |
| arena_text_factuality | 1458 | 1481 |
| Artificial Analysis Coding Index | 47.8 | 78 |
| browsecomp | 37 | 90.8 |
| coding_arena_elo | 1495 | 1662 |
| critpt | 4.6 | 29.1 |
| gdpval | 47.3 | 67.2 |
| GPQA Diamond | 87 | 93.7 |
| HLE | 30.8 | 64.7 |
| HLE (with tools) | 43.4 | 64.7 |
| livebench_agentic_coding | 39.7 | 65.2 |
| livebench_coding | 79.7 | 81.5 |
| livebench_data_analysis | 74.4 | 74.5 |
| livebench_instruction_following | 62.5 | 63.8 |
| livebench_language | 81.3 | 88.7 |
| livebench_math | 90.4 | 95.7 |
| livebench_reasoning | 80.1 | 91.2 |
| MCP Atlas | 69.8 | 85.8 |
| MMMU-Pro | 74 | 84.7 |
| OmniScience Accuracy | 46.6 | 60.9 |
| OmniScience Non-Hallucination | 39 | 40.5 |
| OSWorld-Verified | 66.3 | 4 |
| scicode | 50 | 55.7 |
| simplebench | 62 | 80.6 |
| simpleqa_verified | 41.8 | 56.7 |
| SWE-bench Multilingual | 77.5 | 89.5 |
| SWE-bench Pro | 57.1 | 79.2 |
| SWE-bench Verified | 81.5 | 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.