| Benchmark | Kimi K2.6 | Kimi K2 (Reasoning) |
|---|---|---|
| AA Agentic Index | 58.7 | 47.9 |
| AA Intelligence | 45.1 | 33.5 |
| AA-LCR | 76.7 | 70.3 |
| AA-Omniscience | 6.4 | -21.4 |
| Artificial Analysis Coding Index | 61.8 | 34.8 |
| browsecomp | 86.3 | 60.2 |
| critpt | 8 | 2.6 |
| frontiermath_tier_4 | 14.6 | 0 |
| gdpval | 41.4 | 24.5 |
| GPQA Diamond | 91.1 | 84.5 |
| HLE | 37.5 | 23.9 |
| HLE (with tools) | 54 | 44.9 |
| IFBench | 76 | 68.1 |
| imo_answer_bench | 86 | 78.6 |
| LiveCodeBench | 89.6 | 79.2 |
| LiveCodeBench v6 | 89.6 | 83.1 |
| MMLU-Pro | 87.1 | 84.6 |
| OmniScience Accuracy | 32.8 | 30.9 |
| OmniScience Non-Hallucination | 60.7 | 25.8 |
| scicode | 53.5 | 44.8 |
| SWE-bench Multilingual | 76.7 | 61.1 |
| SWE-bench Verified | 80.2 | 71.3 |
| Terminal-Bench 2.0 | 66.7 | 35.7 |
| Terminal-Bench Hard | 43.9 | 31.1 |
| vectara_answer_rate | 99.7 | 98.6 |
| vectara_avg_summary_length | 116.7 | 59.2 |
| vectara_factual_consistency | 89.2 | 82.1 |
| vectara_hallucination_rate ↓ | 10.8 | 17.9 |
| τ²-Bench Telecom (AA run) | 95.9 | 93 |
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.