| Benchmark | Claude Opus 4.5 | Kimi K3 |
|---|---|---|
| AA Agentic Index | 59.6 | 54.3 |
| AA Intelligence | 41.9 | 60 |
| AA-LCR | 76 | 82.7 |
| AA-Omniscience | 14 | 46 |
| aime_2026 | 95.1 | 96.7 |
| arc_agi_1 | 82 | 65.7 |
| ARC-AGI-2 | 37.6 | 12.4 |
| arena_elo | 1473 | 1489 |
| arena_text_factuality | 1458 | 1471 |
| Artificial Analysis Coding Index | 47.8 | 76.2 |
| browsecomp | 37 | 91.2 |
| charxiv_rq | 68.5 | 84.8 |
| coding_arena_elo | 1495 | 1674 |
| critpt | 4.6 | 23.4 |
| cybergym | 50.6 | 80 |
| deepsearchqa_f1 | 76.1 | 95 |
| gdpval | 47.3 | 59.4 |
| GPQA Diamond | 87 | 93.5 |
| HLE | 30.8 | 56 |
| hmmt_feb_2026 | 85.3 | 97 |
| livebench_agentic_coding | 39.7 | 62.2 |
| livebench_coding | 79.7 | 81.5 |
| livebench_data_analysis | 74.4 | 78.7 |
| livebench_instruction_following | 62.5 | 71.4 |
| livebench_language | 81.3 | 85.5 |
| livebench_math | 90.4 | 84.4 |
| livebench_reasoning | 80.1 | 90.7 |
| mathvision | 77.1 | 97.8 |
| MCP Atlas | 69.8 | 84.2 |
| MMMU-Pro | 74 | 83.4 |
| MMVU | 77.3 | 82.1 |
| OmniScience Accuracy | 46.6 | 47.6 |
| OmniScience Non-Hallucination | 39 | 49.1 |
| OSWorld-Verified | 66.3 | 84.8 |
| scicode | 50 | 58.7 |
| simplebench | 62 | 60.7 |
| simpleqa_verified | 41.8 | 42.7 |
| toolathlon | 43.5 | 73.2 |
| WorldVQA | 36.8 | 51 |
| ZeroBench | 3 | 0.4 |
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.