| Benchmark | GPT-4o mini | Qwen3 VL 4B (Reasoning) |
|---|---|---|
| AA Agentic Index | 1 | 14.4 |
| AA Intelligence | 7 | 7.7 |
| AI2D | 75.2 | 84.9 |
| Artificial Analysis Coding Index | 11.4 | 6.7 |
| BLINK | 51.9 | 65.1 |
| gdpval | 0 | 13.8 |
| GPQA Diamond | 42.6 | 64.1 |
| HLE | 4.2 | 4.6 |
| IFBench | 31 | 36.6 |
| LiveCodeBench | 35.5 | 51.3 |
| MathVista | 56.7 | 79.5 |
| MMMU (val) (Pass@1) | 52.1 | 70.8 |
| MMMU-Pro | 41.5 | 57 |
| scicode | 22.9 | 17.1 |
| Video-MME | 68.9 | 59.7 |
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.