| Benchmark | Gemini 2.5 Flash | Kimi K3 |
|---|---|---|
| AA Agentic Index | 18.8 | 54.3 |
| AA Intelligence | 17.7 | 60 |
| AA-LCR | 65.7 | 82.7 |
| AA-Omniscience | -29.8 | 46 |
| arc_agi_1 | 33.3 | 65.7 |
| ARC-AGI-2 | 2.5 | 12.4 |
| arena_text_factuality | 1437 | 1471 |
| Artificial Analysis Coding Index | 22.2 | 76.2 |
| critpt | 1.4 | 23.4 |
| gdpval | 11.9 | 59.4 |
| GPQA Diamond | 82.8 | 93.5 |
| HLE | 12.1 | 56 |
| MMMU-Pro | 69.1 | 83.4 |
| OmniScience Accuracy | 26.1 | 47.6 |
| OmniScience Non-Hallucination | 24.7 | 49.1 |
| scicode | 35.9 | 58.7 |
| simplebench | 41.2 | 60.7 |
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.