How to Run Flux Dev GGUF Models in ComfyUI (Low VRAM Guide)
Table of Contents
- 1. Introduction to GGUF Models in ComfyUI
- 2. Additional Requirements for Running the Flux Dev GGUF Model in ComfyUI
- 4. Custom Flux Dev GGUF Workflow for ComfyUI + Installing Missing GGUF Nodes
- 5. Flux Dev GGUF Image Generation Settings
- 6. Flux Dev GGUF Model Image Generation Examples
- 7. Conclusion: High-Quality Image Generation Without High-End Hardware
1. Introduction to GGUF Models in ComfyUI
In the world of AI-driven image generation, the choice of model format can significantly impact performance and usability. One of the most notable formats currently gaining traction is the GGUF (Generalized Graphical Unified Format) model, particularly for use in ComfyUI. Unlike the traditional Safetensor models, GGUF models are designed to be more efficient, allowing for smoother operation on systems with limited resources. This is especially important for users with PCs that have only 4–6 GB of RAM. By utilizing GGUF models, users can experience enhanced image generation capabilities without the need for high-end hardware.
For those looking to get started with ComfyUI, it's essential to first set up the environment correctly. If you haven't done so already, check out our guide below:
Next, we need to install some additional files to ensure the GGUF Flux Dev model runs smoothly. Let’s get started!
2. Additional Requirements for Running the Flux Dev GGUF Model in ComfyUI
To fully utilize the Flux Dev GGUF model in ComfyUI, you need to download the required files listed in the table below and place each one in its correct local directory. This step is essential to ensure everything works smoothly and you get the best image generation results.
Required Files, Download Links, and Their Directories
| Model/File Name | Download Link | Directory |
|---|---|---|
| flux1-dev-Q3_K_S.gguf | Download | C:\comfyui\ComfyUI_windows_portable\ComfyUI\models\diffusion_models |
| t5-v1_1-xxl-encoder-Q3_K_S.gguf | Download | C:\comfyui\ComfyUI_windows_portable\ComfyUI\models\clip |
| clip_l.safetensors | Download | C:\comfyui\ComfyUI_windows_portable\ComfyUI\models\clip |
| ae.safetensor | Download | C:\comfyui\ComfyUI_windows_portable\ComfyUI\models\vae |
Be sure to download the correct files and place them in their designated (local) folders to correctly set up the Flux Dev GGUF model in ComfyUI.
Note: Flux Dev GGUF models come in different quantization levels — higher levels need more VRAM. As a rule of thumb, the model file size (in GB) is roughly the VRAM it requires. If you use a different GGUF model than flux1-dev-Q3_K_S.gguf, make sure to also download the matching clip encoder for it.
4. Custom Flux Dev GGUF Workflow for ComfyUI + Installing Missing GGUF Nodes
Once all required models and files are placed in their correct directories, the next step is to add a custom workflow to smoothly use the Flux Dev GGUF model in ComfyUI. You can download the workflow JSON here: Download Workflow JSON. After downloading, simply drag and drop the JSON file onto the ComfyUI canvas to load the workflow.
Installing Missing GGUF Node
If you're new to GGUF models, you might see two red nodes in the workflow — Unet Loader (GGUF) and DualCLIPLoader (GGUF). Don't worry — that just means the nodes are missing. Let's fix that by installing them now.
Make sure to install any missing nodes by following these steps:
-
Click the “Manager” button in the top-right corner of ComfyUI.
-
Select “Install Missing Custom Nodes” from the menu.
-
Find the “ComfyUI-GGUF” node in the list.
-
Choose the latest version and install it.
-
Once installed, restart ComfyUI.
-
After restarting, confirm that the previously missing nodes are now installed and no longer outlined in red.
5. Flux Dev GGUF Image Generation Settings
With all required nodes installed correctly, we can now proceed by selecting the correct models you downloaded earlier. In the workflow, select the following nodes and assign these files:
-
-
Unet Loader (GGUF): load flux1-dev-Q3_K_S.gguf
-
DualCLIPLoader (GGUF): load t5-v1_1-xxl-encoder-Q3_K_S.gguf and clip_l.safetensors
-
Load VAE: load ae.safetensor
| Setting | Value |
|---|---|
| Prompt | anime woman, sensual pose, pink hair, black dress, cleavage |
| Width | 1024 |
| Height | 1024 |
| Steps | 20 |
| Sampler | Euler |
| Scheduler | Simple |
Click Generate and watch your image come to life — in the next section, we’ll showcase more examples using different prompts, styles, and settings.
6. Flux Dev GGUF Model Image Generation Examples
Below are some examples of images generated using the Flux Dev model, showcasing its capabilities across styles and formats.
Realistic Portrait
This prompt highlights the model’s ability to produce soft, photorealistic lighting and skin textures.
| Setting | Value |
|---|---|
| Prompt | Latina woman, upper body portrait, elegant black lingerie, soft bedroom lighting, warm skin tones, long wavy dark hair, natural makeup, subtle freckles, seductive gaze, slight smile, shallow depth of field, cleavage visible |
| Width | 1280 |
| Height | 720 |
| Steps | 20 |
| Sampler | Euler |
| Scheduler | Simple |
Anime Style
This anime-themed prompt showcases the model’s creative control and dramatic aesthetic.
| Setting | Value |
|---|---|
| Prompt | anime girl in a black silk dress, long flowing purple hair, soft glow, starry background, standing gracefully, cleavage emphasized |
| Width | 1024 |
| Height | 1024 |
| Steps | 30 |
| Sampler | Euler |
| Scheduler | Simple |
Cartoon Pin-Up
A stylized cartoon-style render, bold and vibrant with a vertical layout ideal for mobile display.
| Setting | Value |
|---|---|
| Prompt | cartoon-style pin-up girl in tight red lingerie, big eyes, confident pose, neon lights background, exaggerated curves, cleavage visible |
| Width | 720 |
| Height | 1280 |
| Steps | 35 |
| Sampler | Euler |
| Scheduler | Simple |
7. Conclusion: High-Quality Image Generation Without High-End Hardware
Running the Flux Dev GGUF model in ComfyUI opens the door to powerful image generation — even on low-end PCs with limited VRAM. With the right setup, proper node installation, and lightweight GGUF models, you can enjoy creative freedom without needing a dedicated GPU. Whether you're working with 4–6 GB of RAM or just want a more efficient local workflow, this guide shows that high performance isn't just for high-spec machines.
Now that you're up and running, try experimenting with different prompts, styles, and GGUF quantization levels to find the balance that works best for your system. And remember — the ComfyUI community continues to grow, so there are always new tools, nodes, and workflows to explore.
Happy generating!
