# gguf

A ComfyUI custom node that adds support for loading GGUF-format quantized diffusion models directly in ComfyUI, covering image, video and audio models with a range of quantization levels and no extra Python dependencies.

This node lets ComfyUI load models stored in the GGUF quantization format, a format more commonly associated with large language model inference, extending it to image, video and audio diffusion models. It installs either as a full downloadable ComfyUI pack that can be decompressed and run directly, or by cloning the node's own repository into ComfyUI's custom_nodes folder for an existing installation, and can also be installed through comfy-cli or the ComfyUI manager's node search.

Bundled features include a tensor cutter and tensor booster for adjusting a loaded model's tensors, a GGUF file converter, and support for the s, k, t, i, mxfp4 and nvfp4 quantization families. The project notes that it shares part of its code base with the separate, similarly named comfyui-gguf node, crediting that project as an inspiration, but describes its own engine as natively coded rather than a repackaging, and states the two nodes can coexist without conflict.

Website: https://github.com/calcuis/gguf