| Benchmark | GPT-4o mini | Kimi K2.5 |
|---|---|---|
| AA Agentic Index | 1 | 52.8 |
| AA Intelligence | 7 | 36 |
| ARC-AGI-2 | 0 | 11.8 |
| Artificial Analysis Coding Index | 11.4 | 46.8 |
| Fortress | 48.1 | 41.1 |
| gdpval | 0 | 38.3 |
| GPQA Diamond | 42.6 | 87.9 |
| HLE | 4.2 | 50.2 |
| IFBench | 31 | 70.2 |
| MathVista | 56.7 | 90.1 |
| MMMU-Pro | 41.5 | 78.5 |
| scicode | 22.9 | 49 |
| simplebench | 10.7 | 46.8 |
| SWE-bench Verified | 8.7 | 76.8 |
| TauBench V3 - Banking | 2.9 | 14.2 |
| Terminal-Bench 2.1 | 5.6 | 45.7 |
| Video-MME | 68.9 | 87.4 |
| τ³-Bench | 2.9 | 66 |
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.