Category ranking · 7 models · checked 2026-10-13
Vision-language
Ranked by MMMU (val). Source: Model cards (vendor-reported).
MMMU (val)
3 with a published score, 4 not yet measured.
# | Model | Parameters | Score | Source | Licence |
|---|---|---|---|---|---|
| 1 | Qwen2.5 VL 7B Alibaba Qwen | 8.3 B | 58.6%Sourced | Source MMMU val, model card table. | Apache-2.0 |
| 2 | Kimi-VL A3B Moonshot AI | 16.4 B | 57%Sourced | Source MMMU-Val pass@1, model card table. | MIT |
| 3 | SmolVLM2 2.2B Hugging Face | 2.2 B | 42%Sourced | Source MMMU column, model card table. | Apache-2.0 |
| — | Gemma 3 12B Google | 12.2 B | Not yet measured | Model card is gated; not read. | Gemma Terms of Use |
| — | InternVL3 8B Shanghai AI Lab | 7.9 B | Not yet measured | — | Apache-2.0 |
| — | Qwen3 VL 30B-A3B Alibaba Qwen | 31.1 B | Not yet measured | Scores published as images only. | Apache-2.0 |
| — | Qwen3 VL 8B Alibaba Qwen | 8.8 B | Not yet measured | Scores published as images only. | Apache-2.0 |
How to read this: one public benchmark per category, so you can compare like with like. We copy the published number and link it; we do not re-run it, average different benchmarks or fill gaps. Vendor-reported numbers use each vendor's own settings, so treat small gaps as ties. “Not yet measured” means we found no comparable public score, not that the model is weak.
Every model was checked against the Hugging Face API on 2026-10-13: it exists, is public, and has the parameter count and licence shown. Other categories: Math and reasoning · Coding · OCR and documents · Speech to text · Text to speech · Image generation · Embeddings and rerankers · Small fast 'System 1' · Biomedical · Tool use and agents · Multilingual and translation · Time series · Robotics (VLA).
Machine-readable: /api/v1/categories/vision-language.json