Vision GGUF models
17604 GGUF models: models that read images through an mmproj file. Sizes run from 1M to 2780B parameters. The longest trained context is 10240K tokens.
About Vision GGUF models
17,604 GGUF models are filed under Vision: models that read images through an mmproj file. The most downloaded are Qwen3.8 Flash Next GSQ RCO, Qwen3.8-27B and Ornith-1.5-9B. Parameter counts run from 1M to 2780B, and the longest trained context window is 10240K tokens.
Vision models usually ship an extra mmproj file next to the main model. You need both, and a runtime with multimodal support such as llama.cpp’s multimodal tools. The families you will see most are Qwen (6684), Gemma (2948) and Mistral (336).
Frequently asked questions
What are Vision GGUF models?
Models that read images through an mmproj file. There are 17,604 of them on this site.
Which Vision GGUF model should I start with?
The most downloaded is Qwen3.8 Flash Next GSQ RCO. Choose the size that fits your GPU, then a Q4_K_M quant unless you have memory to spare.