| Benchmark | Kimi K2 (Reasoning) | o3 |
|---|---|---|
| AA Agentic Index | 47.9 | 36.1 |
| AA Intelligence | 33.5 | 31.1 |
| AA-LCR | 70.3 | 73.3 |
| AA-Omniscience | -21.4 | -15.3 |
| aider_polyglot | 59.1 | 81.3 |
| AIME 2025 | 94.5 | 89.2 |
| Artificial Analysis Coding Index | 34.8 | 38.4 |
| browsecomp | 60.2 | 49.7 |
| critpt | 2.6 | 1.1 |
| frontiermath_tier_4 | 0 | 2.1 |
| gdpval | 24.5 | 12.8 |
| GPQA Diamond | 84.5 | 83.3 |
| HLE | 23.9 | 20.3 |
| hmmt_feb_2025 | 93.3 | 77.5 |
| IFBench | 68.1 | 71.4 |
| LiveCodeBench | 79.2 | 84.7 |
| longbench_v2 | 45.1 | 58.8 |
| MMLU-Pro | 84.6 | 85 |
| multichallenge | 66.4 | 60.4 |
| OmniScience Accuracy | 30.9 | 38.6 |
| OmniScience Non-Hallucination | 25.8 | 12.9 |
| scicode | 44.8 | 41 |
| simplebench | 39.6 | 53.1 |
| swe_bench_bash | 63.4 | 58.4 |
| SWE-bench Verified | 71.3 | 69.1 |
| Terminal-Bench Hard | 31.1 | 37.1 |
| τ²-Bench Telecom (AA run) | 93 | 80.7 |
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.