Run LLM/VLM models natively in ComfyUI based on llama.cpp
[📃中文版]

• Initial release, added support for Qwen3-VL

cd ComfyUI/custom_nodes
git clone https://github.com/lihaoyun6/ComfyUI-llama-cpp.git
python -m pip install -r ComfyUI-llama-cpp/requirements.txt

• Install a prebuilt wheel from https://github.com/JamePeng/llama-cpp-python/releases, or build it from source according to your system.

• Place your model files in the ComfyUI/models/LLM folder.

If you need a VLM model to process image input, don't forget to download the mmproj weights.

• llama-cpp-python @JamePeng
• ComfyUI-llama-cpp @kijai
• ComfyUI @comfyanonymous