| Benchmark | Kimi K2 (Reasoning) | Opus 5 |
|---|---|---|
| AA Agentic Index | 47.9 | 59.2 |
| AA Intelligence | 33.5 | 63.1 |
| AA-LCR | 70.3 | 78.7 |
| AA-Omniscience | -21.4 | 37.1 |
| Artificial Analysis Coding Index | 34.8 | 78 |
| browsecomp | 60.2 | 90.8 |
| critpt | 2.6 | 29.1 |
| gdpval | 24.5 | 67.2 |
| GPQA Diamond | 84.5 | 93.7 |
| HealthBench | 58 | 67.1 |
| HLE | 23.9 | 64.7 |
| HLE (with tools) | 44.9 | 64.7 |
| OmniScience Accuracy | 30.9 | 60.9 |
| OmniScience Non-Hallucination | 25.8 | 40.5 |
| scicode | 44.8 | 55.7 |
| simplebench | 39.6 | 80.6 |
| SWE-bench Multilingual | 61.1 | 89.5 |
| SWE-bench Verified | 71.3 | 96 |
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.