| Benchmark | Kimi K2.5 | Llama 3.1 Instruct 405B |
|---|---|---|
| AA Agentic Index | 52.8 | 6.3 |
| AA Intelligence | 36 | 8.5 |
| AA-LCR | 73 | 24.7 |
| AA-Omniscience | -7.3 | -17.1 |
| Artificial Analysis Coding Index | 46.8 | 14.5 |
| critpt | 3.1 | 0 |
| Fortress | 41.1 | 20.6 |
| gdpval | 38.3 | 0 |
| GPQA Diamond | 87.9 | 51.5 |
| HLE | 50.2 | 4.2 |
| IFBench | 70.2 | 39 |
| longbench_v2 | 61 | 36.1 |
| MMLU-Pro | 87.1 | 73.4 |
| OmniScience Accuracy | 35.2 | 23.2 |
| OmniScience Non-Hallucination | 50 | 49 |
| scicode | 49 | 29.9 |
| simplebench | 46.8 | 23 |
| SWE-bench Pro | 53.8 | 11.2 |
| SWE-bench Verified | 76.8 | 24.5 |
| Terminal-Bench Hard | 34.8 | 6.8 |
| τ²-Bench Telecom (AA run) | 95.9 | 19 |
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.