Hugging Face Top Multimodal Models
google/gemma-4-31B-it tops the Multimodal Models chart with 8,301,351 downloads. Qwen/Qwen3.8-27B is second with 7,892,866 downloads.
Multimodal AI is 2026's dominant model category: a single system that understands and generates text, images, audio and video. This chart ranks the most-downloaded multimodal and vision-language models on Hugging Face by trailing-30-day downloads — the clearest signal of what practitioners actually build with.
ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF
7,668,149 downloads in the trailing 30 days. 1,128 likes.
Qwen/Qwen3-VL-8B-Instruct
6,902,450 downloads in the trailing 30 days. 1,235 likes.
google/gemma-4-26B-A4B-it
6,032,823 downloads in the trailing 30 days. 1,724 likes.
Qwen/Qwen2.5-VL-7B-Instruct
5,838,999 downloads in the trailing 30 days. 1,762 likes.
microsoft/Florence-2-base
5,241,149 downloads in the trailing 30 days. 433 likes.
Editor’s Notes
- Ranked by trailing-30-day downloads from the Hugging Face Hub API, refreshed daily.
- Covers multimodal / vision-language models (pipeline tag image-text-to-text and vision variants).
- Download counts are request counts to repository files and are not deduplicated to unique users.
Last updated: 2026-11-02