Across 12 shared benchmarks, Gemma 3 27B scores higher on 9 and Qwen2.5 Omni 7B on 3. The widest gap is mathvision, where Gemma 3 27B scores 46 against 25. Gemma 3 27B is the cheaper of the two on tracked API pricing ($0.08 against $0.10 per million input tokens).
| Benchmark | Gemma 3 27B | Qwen2.5 Omni 7B |
|---|---|---|
| AI2D | 84.5 | 83.2 |
| chartqa | 78 | 85.3 |
| docvqa | 86.6 | 95.2 |
| GPQA Diamond | 46 | 30.8 |
| GSM8K | 95.9 | 88.7 |
| humaneval | 87.8 | 78.7 |
| mathvision | 46 | 25 |
| MathVista | 67.6 | 67.9 |
| MBPP | 74.4 | 0.7 |
| MMLU-Pro | 67.6 | 47 |
| MMMU | 64.9 | 59.2 |
| MMMU-Pro | 49.7 | 36.6 |
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. Quoted rates are the price-setter row we currently track for each model — its direct or vendor-official listing where one exists (deepinfra, alibaba-official), otherwise the lowest tracked offer.