| Benchmark | Claude Sonnet 4.6 | Kimi K2 (Reasoning) |
|---|---|---|
| AA Agentic Index | 61.6 | 47.9 |
| AA Intelligence | 48.4 | 33.5 |
| AA-LCR | 74 | 70.3 |
| AA-Omniscience | 12.2 | -21.4 |
| Artificial Analysis Coding Index | 63 | 34.8 |
| browsecomp | 76.2 | 60.2 |
| critpt | 3.1 | 2.6 |
| frontiermath_tier_4 | 8.3 | 0 |
| gdpval | 54.8 | 24.5 |
| GPQA (unspecified) | 89.9 | 74.2 |
| GPQA Diamond | 89.9 | 84.5 |
| HLE | 49 | 23.9 |
| HLE (with tools) | 46.8 | 44.9 |
| IFBench | 56.6 | 68.1 |
| OmniScience Accuracy | 40.9 | 30.9 |
| OmniScience Non-Hallucination | 51.6 | 25.8 |
| scicode | 47 | 44.8 |
| SWE-bench Verified | 79.6 | 71.3 |
| Terminal-Bench 2.0 | 59.1 | 35.7 |
| Terminal-Bench Hard | 59.1 | 31.1 |
| vectara_answer_rate | 99.9 | 98.6 |
| vectara_avg_summary_length | 114.7 | 59.2 |
| vectara_factual_consistency | 89.4 | 82.1 |
| vectara_hallucination_rate ↓ | 10.6 | 17.9 |
| τ²-Bench Telecom (AA run) | 97.9 | 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.