| Benchmark | Claude Opus 4.6 | Kimi K3 |
|---|---|---|
| AA Agentic Index | 67.6 | 54.3 |
| AA Intelligence | 44.9 | 60 |
| AA-LCR | 74.3 | 82.7 |
| AA-Omniscience | 13.7 | 46 |
| aime_2026 | 96.7 | 96.7 |
| apexAgents | 33 | 41.3 |
| arc_agi_1 | 93 | 65.7 |
| ARC-AGI-2 | 68.8 | 12.4 |
| arena_elo | 1497 | 1489 |
| arena_text_factuality | 1487 | 1471 |
| Artificial Analysis Coding Index | 48.1 | 76.2 |
| baby_vision | 14.8 | 85.7 |
| baby_vision_with_python | 38.4 | 85.7 |
| browsecomp | 86.8 | 91.2 |
| charxiv_rq | 69.1 | 84.8 |
| coding_arena_elo | 1545 | 1674 |
| critpt | 12.6 | 23.4 |
| cybergym | 73.8 | 80 |
| deepsearchqa_f1 | 91.3 | 95 |
| gdpval | 55.9 | 59.4 |
| GPQA Diamond | 91.3 | 93.5 |
| HLE | 53.1 | 56 |
| hmmt_feb_2026 | 96.2 | 97 |
| livebench_agentic_coding | 49 | 62.2 |
| livebench_coding | 78.2 | 81.5 |
| livebench_data_analysis | 69.9 | 78.7 |
| livebench_instruction_following | 63.3 | 71.4 |
| livebench_language | 83.3 | 85.5 |
| livebench_math | 89.3 | 84.4 |
| livebench_reasoning | 88.7 | 90.7 |
| matharena_visual_math_overall | 72.3 | 91.3 |
| mathvision | 71.2 | 97.8 |
| MCP Atlas | 76.8 | 84.2 |
| MMMU-Pro | 77.3 | 83.4 |
| officeqa_pro | 57.1 | 63.3 |
| OmniScience Accuracy | 47 | 47.6 |
| OmniScience Non-Hallucination | 37.2 | 49.1 |
| OSWorld-Verified | 72.7 | 84.8 |
| scicode | 52 | 58.7 |
| simplebench | 67.6 | 60.7 |
| simpleqa_verified | 46.5 | 42.7 |
| toolathlon | 56.8 | 73.2 |
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.