| Benchmark | DeepSeek-R1 | Kimi K2 (Reasoning) |
|---|---|---|
| AA Agentic Index | 3.1 | 47.9 |
| AA Intelligence | 18.6 | 33.5 |
| AA-LCR | 56 | 70.3 |
| AA-Omniscience | -31.3 | -21.4 |
| aider_polyglot | 71.6 | 59.1 |
| AIME 2025 | 87.5 | 94.5 |
| Artificial Analysis Coding Index | 24.6 | 34.8 |
| browsecomp | 8.9 | 60.2 |
| browsecomp_zh | 35.7 | 62.3 |
| critpt | 0.6 | 2.6 |
| gdpval | 1.5 | 24.5 |
| GPQA Diamond | 81 | 84.5 |
| HLE | 17.7 | 23.9 |
| HMMT 2025 | 79.4 | 89.4 |
| hmmt_feb_2025 | 76.7 | 93.3 |
| IFBench | 39 | 68.1 |
| LiveCodeBench | 84.4 | 79.2 |
| longbench_v2 | 58.3 | 45.1 |
| mmlu_redux | 93.4 | 94.4 |
| MMLU-Pro | 85 | 84.6 |
| multichallenge | 45 | 66.4 |
| OmniScience Accuracy | 30.7 | 30.9 |
| OmniScience Non-Hallucination | 10.5 | 25.8 |
| scicode | 35.7 | 44.8 |
| simplebench | 40.8 | 39.6 |
| SWE-bench Multilingual | 30.5 | 61.1 |
| SWE-bench Verified | 57.6 | 71.3 |
| Terminal-Bench | 5.7 | 47.1 |
| Terminal-Bench Hard | 6.1 | 31.1 |
| vectara_answer_rate | 97 | 98.6 |
| vectara_avg_summary_length | 93.5 | 59.2 |
| vectara_factual_consistency | 88.7 | 82.1 |
| vectara_hallucination_rate ↓ | 11.3 | 17.9 |
| τ²-Bench Telecom (AA run) | 11.4 | 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.