| Benchmark | Kimi K2.6 | Kimi K3 |
|---|---|---|
| AA Agentic Index | 58.7 | 54.3 |
| AA Intelligence | 45.1 | 60 |
| AA-LCR | 76.7 | 82.7 |
| AA-Omniscience | 6.4 | 46 |
| aime_2026 | 96.4 | 96.7 |
| apex_agents | 27.9 | 41 |
| apexAgents | 28.5 | 41.3 |
| arena_elo | 1461 | 1489 |
| arena_text_factuality | 1455 | 1471 |
| Artificial Analysis Coding Index | 61.8 | 76.2 |
| baby_vision | 68.5 | 85.7 |
| baby_vision_with_python | 68.5 | 85.7 |
| browsecomp | 86.3 | 91.2 |
| charxiv_rq | 86.7 | 84.8 |
| coding_arena_elo | 1509 | 1674 |
| critpt | 8 | 23.4 |
| deepsearchqa_f1 | 92.5 | 95 |
| Finance Agent v2 | 44.9 | 54.4 |
| gdpval | 41.4 | 59.4 |
| GDPval-AA v2 | 1190 | 1687 |
| GPQA Diamond | 91.1 | 93.5 |
| HLE | 37.5 | 56 |
| hmmt_feb_2026 | 94.7 | 97 |
| itbenchSre | 31.2 | 47.7 |
| livebench_agentic_coding | 46.9 | 62.2 |
| livebench_coding | 78.6 | 81.5 |
| livebench_data_analysis | 65.1 | 78.7 |
| livebench_instruction_following | 64.4 | 71.4 |
| livebench_language | 75.1 | 85.5 |
| livebench_math | 84.3 | 84.4 |
| livebench_reasoning | 79.4 | 90.7 |
| mathvision | 93.2 | 97.8 |
| MCP Atlas | 68.1 | 84.2 |
| MMMU-Pro | 80.1 | 83.4 |
| OmniScience Accuracy | 32.8 | 47.6 |
| OmniScience Non-Hallucination | 60.7 | 49.1 |
| OSWorld-Verified | 73.1 | 84.8 |
| scicode | 53.5 | 58.7 |
| simpleqa_verified | 38.7 | 42.7 |
| Tau 3 Banking | 20.6 | 33.4 |
| TauBench V3 - Banking | 23.3 | 46 |
| Terminal-Bench 2.1 | 65.9 | 88.3 |
| toolathlon | 50 | 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.