Across 12 shared benchmarks, Gemini 1.5 Pro scores higher on 7 and Qwen2.5 VL 32B Instruct on 5. The widest gap is GPQA Diamond, where Gemini 1.5 Pro scores 59.1 against 46.
| Benchmark | Gemini 1.5 Pro | Qwen2.5 VL 32B Instruct |
|---|---|---|
| arena_vision | 1179 | 1153 |
| docvqa | 91.5 | 94.8 |
| GPQA Diamond | 59.1 | 46 |
| humaneval | 86.6 | 91.5 |
| MATH | 92 | 82.2 |
| MathVista | 70.6 | 74.7 |
| mmlu | 86.8 | 78.4 |
| MMLU-Pro | 75.8 | 68.8 |
| MMMU | 68.4 | 70 |
| MMMU-Pro | 55 | 49.5 |
| ocrbench_v2 | 51.6 | 59.1 |
| Video-MME | 78.6 | 77.9 |
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.