| Benchmark | Gemini 1.5 Pro | Qwen3 VL 4B (Reasoning) |
|---|---|---|
| AA Intelligence | 10 | 7.7 |
| AI2D | 80.9 | 84.9 |
| Artificial Analysis Coding Index | 23.6 | 6.7 |
| BLINK | 61 | 65.1 |
| chartqa | 88.7 | 84 |
| GPQA Diamond | 59.1 | 64.1 |
| HLE | 4.9 | 4.6 |
| ifeval | 89.4 | 82.6 |
| LiveCodeBench | 30.5 | 51.3 |
| MathVista | 70.6 | 79.5 |
| MMLU-Pro | 75.8 | 73.6 |
| MMMU (val) (Pass@1) | 54.1 | 70.8 |
| MMMU-Pro | 55 | 57 |
| OCRBench | 800 | 81.6 |
| scicode | 29.5 | 17.1 |
| Video-MME | 78.6 | 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.