Across 11 shared benchmarks, Gemini 1.5 Pro scores higher on 8 and Qwen2 VL 72B Instruct on 3. The widest gap is MMMU (val) (Pass@1), where Qwen2 VL 72B Instruct scores 64.5 against 54.1.
| Benchmark | Gemini 1.5 Pro | Qwen2 VL 72B Instruct |
|---|---|---|
| chartqa | 88.7 | 88.3 |
| docvqa | 91.5 | 96.5 |
| MathVista | 70.6 | 70.5 |
| MMMU | 68.4 | 64.5 |
| MMMU (val) (Pass@1) | 54.1 | 64.5 |
| MMMU-Pro | 55 | 46.2 |
| OCRBench | 800 | 877 |
| ocrbench_v2 | 51.6 | 46.1 |
| Video-MME | 78.6 | 77.8 |
| Video-MME (w subs) | 81.3 | 77.8 |
| Video-MME (wo subs) | 75 | 71.2 |
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.