| Benchmark | Kimi K3 | Llama 3.1 Instruct 405B |
|---|---|---|
| AA Agentic Index | 54.3 | 6.3 |
| AA Intelligence | 60 | 8.5 |
| AA-LCR | 82.7 | 24.7 |
| AA-Omniscience | 46 | -17.1 |
| Artificial Analysis Coding Index | 76.2 | 14.5 |
| critpt | 23.4 | 0 |
| gdpval | 59.4 | 0 |
| GPQA Diamond | 93.5 | 51.5 |
| HLE | 56 | 4.2 |
| OmniScience Accuracy | 47.6 | 23.2 |
| OmniScience Non-Hallucination | 49.1 | 49 |
| scicode | 58.7 | 29.9 |
| simplebench | 60.7 | 23 |
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.