| Benchmark | Kimi K2 (Reasoning) | Qwen3.5 27B |
|---|---|---|
| AA Agentic Index | 47.9 | 54.6 |
| AA Intelligence | 33.5 | 34.6 |
| AA-LCR | 70.3 | 72.3 |
| AA-Omniscience | -21.4 | -44 |
| Artificial Analysis Coding Index | 34.8 | 34.9 |
| browsecomp | 60.2 | 61 |
| browsecomp_zh | 62.3 | 62.1 |
| critpt | 2.6 | 0.9 |
| gdpval | 24.5 | 33 |
| GPQA Diamond | 84.5 | 85.8 |
| HLE | 23.9 | 48.5 |
| HMMT 2025 | 89.4 | 89.8 |
| hmmt_feb_2025 | 93.3 | 92 |
| hmmt_nov_2025 | 89.2 | 89.8 |
| IFBench | 68.1 | 76.5 |
| imo_answer_bench | 78.6 | 79.9 |
| LiveCodeBench v6 | 83.1 | 80.7 |
| longbench_v2 | 45.1 | 60.6 |
| mmlu_redux | 94.4 | 93.2 |
| MMLU-Pro | 84.6 | 86.1 |
| multichallenge | 66.4 | 60.8 |
| OmniScience Accuracy | 30.9 | 20.7 |
| OmniScience Non-Hallucination | 25.8 | 24.9 |
| scicode | 44.8 | 39.5 |
| Seal-0 | 56.3 | 47.2 |
| SWE-bench Multilingual | 61.1 | 69.3 |
| SWE-bench Verified | 71.3 | 75 |
| Terminal-Bench 2.0 | 35.7 | 41.6 |
| Terminal-Bench Hard | 31.1 | 32.6 |
| vectara_answer_rate | 98.6 | 99.8 |
| vectara_avg_summary_length | 59.2 | 94.4 |
| vectara_factual_consistency | 82.1 | 87.9 |
| vectara_hallucination_rate ↓ | 17.9 | 12.1 |
| τ²-Bench Telecom (AA run) | 93 | 93.9 |
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.