| Benchmark | DeepSeek-R1 | Kimi K2.6 |
|---|---|---|
| AA Agentic Index | 3.1 | 58.7 |
| AA Intelligence | 18.6 | 45.1 |
| AA-LCR | 56 | 76.7 |
| AA-Omniscience | -31.3 | 6.4 |
| Artificial Analysis Coding Index | 24.6 | 61.8 |
| browsecomp | 8.9 | 86.3 |
| Chinese SimpleQA (C-SimpleQA) | 63.7 | 75.9 |
| critpt | 0.6 | 8 |
| gdpval | 1.5 | 41.4 |
| GPQA Diamond | 81 | 91.1 |
| HLE | 17.7 | 37.5 |
| IFBench | 39 | 76 |
| LiveCodeBench | 84.4 | 89.6 |
| MMLU-Pro | 85 | 87.1 |
| OmniScience Accuracy | 30.7 | 32.8 |
| OmniScience Non-Hallucination | 10.5 | 60.7 |
| scicode | 35.7 | 53.5 |
| simpleqa_verified | 27.4 | 38.7 |
| SWE-bench Multilingual | 30.5 | 76.7 |
| SWE-bench Verified | 57.6 | 80.2 |
| TauBench V3 - Banking | 6.4 | 23.3 |
| Terminal-Bench 2.1 | 19.1 | 65.9 |
| Terminal-Bench Hard | 6.1 | 43.9 |
| vectara_answer_rate | 97 | 99.7 |
| vectara_avg_summary_length | 93.5 | 116.7 |
| vectara_factual_consistency | 88.7 | 89.2 |
| vectara_hallucination_rate ↓ | 11.3 | 10.8 |
| τ²-Bench Telecom (AA run) | 11.4 | 95.9 |
| τ³-Bench | 6.4 | 20.6 |
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.