ComfyUI GGUF: How to Install, Load, and Troubleshoot GGUF Models

Use GGUF models in ComfyUI with the ComfyUI-GGUF custom node, Unet Loader GGUF, CLIPLoader GGUF, and the right model folders.

GGUF models in ComfyUI usually need the ComfyUI-GGUF custom node.

The short version: install the custom node, put diffusion model .gguf files in ComfyUI/models/unet or ComfyUI/models/diffusion_models, then replace the normal diffusion model loader with Unet Loader (GGUF).

Quick Answer

Goal	What to Do
Load a Flux GGUF model	Install ComfyUI-GGUF, then use `Unet Loader (GGUF)`
Find the right folder	Start with `ComfyUI/models/unet` or `ComfyUI/models/diffusion_models`
Load a GGUF T5 text encoder	Use a `CLIPLoader (GGUF)` node when the workflow calls for it
Fix missing GGUF loader nodes	Check that `ComfyUI-GGUF` is installed under `custom_nodes` and restart
Fix empty model dropdown	Confirm the file extension, folder, and active ComfyUI environment

What GGUF Means in ComfyUI

GGUF is a model file format commonly used for quantized models. In ComfyUI, it is often used for large transformer or DiT-based diffusion models such as Flux-style workflows.

The reason people use GGUF is practical: quantized files can reduce VRAM requirements compared with full precision model files. The tradeoff is that setup becomes more specific. You need the right custom node and the right loader.

Step 1: Install ComfyUI-GGUF

For a manual install, open a terminal in your ComfyUI folder and clone the custom node:

cd ComfyUI/custom_nodes
git clone https://github.com/city96/ComfyUI-GGUF

Then install its requirements in the same Python environment used by ComfyUI:

cd ComfyUI-GGUF
python -m pip install -r requirements.txt

For a Windows portable package, run from the portable root:

git clone https://github.com/city96/ComfyUI-GGUF ComfyUI\custom_nodes\ComfyUI-GGUF
.\python_embeded\python.exe -s -m pip install -r .\ComfyUI\custom_nodes\ComfyUI-GGUF\requirements.txt

Restart ComfyUI after installing the node.

Install into the right Python

If you use portable ComfyUI, system pip is usually the wrong pip. Use python_embeded\python.exe from the portable package.

Step 2: Put GGUF Files in the Right Folder

The ComfyUI-GGUF README says to place .gguf model files in:

ComfyUI/models/unet

Current ComfyUI folder configuration also maps legacy unet paths to:

ComfyUI/models/diffusion_models

So if one guide says models/unet and another says models/diffusion_models, they may both be describing the same model type. The important part is that GGUF diffusion models are not normal checkpoint files.

Do not put a Flux GGUF diffusion model here:

ComfyUI/models/checkpoints

That folder is for regular checkpoint loader workflows.

For a broader folder map, see Where to Put Safetensors in ComfyUI.

Step 3: Use Unet Loader GGUF

After installing the custom node, search for a node like:

Unet Loader (GGUF)

Use it in place of the normal Load Diffusion Model node when the workflow expects a GGUF diffusion model.

This is the most common fix for searches like:

comfyui unet loader gguf
unet loader gguf comfyui
comfyui gguf loader
comfyui load gguf model

If the node does not appear, ComfyUI-GGUF did not load correctly. Check the startup terminal for an import error.

Step 4: Know When You Need CLIPLoader GGUF

Some GGUF workflows also use quantized text encoders, such as T5 GGUF files.

In that case, the workflow may use a GGUF CLIP loader node instead of the normal CLIP loader. The ComfyUI-GGUF README notes that its CLIP loader can handle GGUF and regular text encoder files.

If your workflow has a missing CLIP loader node:

Confirm ComfyUI-GGUF is installed.
Confirm the text encoder file is in the expected text encoder folder.
Confirm the workflow actually expects a GGUF text encoder, not just a GGUF diffusion model.

Step 5: Restart and Check the Terminal

After installing ComfyUI-GGUF and moving files, restart ComfyUI.

If the loader node is still missing, check the first real error in the terminal. Look for:

IMPORT FAILED
ModuleNotFoundError
No module named 'gguf'

If gguf is missing, install the custom node requirements in the correct Python environment.

Common Problems

Problem	Cause	Fix
`Unet Loader (GGUF)` is missing	Custom node did not load	Reinstall ComfyUI-GGUF and restart
GGUF file not in dropdown	File is in the wrong model folder	Move it to `models/unet` or `models/diffusion_models`
Workflow uses normal Load Checkpoint	Wrong loader for GGUF diffusion model	Replace it with `Unet Loader (GGUF)`
`No module named 'gguf'`	Dependency missing	Install ComfyUI-GGUF requirements
Out of memory still happens	Quantization helps, but does not make every workflow fit	Use a smaller quant, lower resolution, or smaller workflow

GGUF vs Checkpoint vs Safetensors

Format	Typical Folder	Typical Use
`.safetensors` checkpoint	`models/checkpoints`	SD 1.5, SDXL, many all-in-one models
`.safetensors` diffusion model	`models/diffusion_models`	Flux and newer split-model workflows
`.gguf` diffusion model	`models/unet` or `models/diffusion_models`	Quantized Flux or other transformer models
`.gguf` text encoder	text encoder or clip folder, depending on workflow	Quantized T5 or text encoder loaders

The file extension alone is not enough. The workflow and loader node decide where the file belongs.

When GGUF Is a Good Choice

GGUF is worth trying when:

your GPU is short on VRAM
the workflow specifically asks for a GGUF model
you are running Flux-style or other large transformer models
a community workflow provides a tested GGUF setup

It is not the best first choice when:

you are still learning basic checkpoint workflows
the custom node fails to import
you have enough VRAM for a standard model
the workflow instructions are unclear or mix several model formats

How Wonderful Launcher Helps

GGUF setup can touch custom nodes, model folders, Python packages, and workflow loaders at the same time. Wonderful Launcher helps when you want to:

keep a stable non-GGUF environment separate from a GGUF experiment
preserve working workflows before adding custom nodes
collect startup logs when loader nodes fail
recover models and workflows before rebuilding ComfyUI

If GGUF install breaks other nodes, start with ComfyUI Dependency Conflicts.