CoolFace
Modelpublic

Abiray/Qwen-Image-2.1-GGUF

sourceHugging Faceotherupdated 6d agoView on Hugging Face
92likes47kdownloads
Model Card

Qwen-Image-2.1 GGUF Quants

This repository provides quantized GGUF checkpoints for Qwen/Qwen-Image-2.1, optimized for low-VRAM inference in ComfyUI using the ComfyUI-GGUF custom node.

Quantizing the diffusion transformer drastically reduces memory pressure during generation while preserving sharp detail, composition, and prompt alignment.

📦 Quantization Breakdown & File Details

FilenameQuant TypeSizeRecommended VRAM / Profile
qwen_image_2.1_Q8_0.ggufQ8_07.59 GB12 GB+ (Near-lossless fidelity)
qwen_image_2.1_Q6_K.ggufQ6_K5.88 GB10–12 GB (High quality sweet spot)
qwen_image_2.1_Q5_K_M.ggufQ5_K_M5.01 GB8–10 GB (Balanced performance & memory)
qwen_image_2.1_Q4_K_M.ggufQ4_K_M4.19 GB6–8 GB (Standard consumer GPU baseline)
qwen_image_2.1_Q4_K_S.ggufQ4_K_S4.06 GB6–8 GB (Compact 4-bit)
qwen_image_2.1_Q3_K_M.ggufQ3_K_M3.19 GB4–6 GB (Extreme budget / minimum VRAM)

🖼️ Sample Generations (Q4KM)

<Gallery />

🧩 Required Components (Text Encoders & VAE)

To run these models in ComfyUI, you will need the matching text encoders and VAE:

  1. 1.Text Encoders:
  2. 2.Download from: Comfy-Org/Qwen-Image-2.1 Text Encoders
  3. 3.Place files into: ComfyUI/models/clip/ (or ComfyUI/models/text_encoders/)
  1. 1.VAE:
  2. 2.Download from: Comfy-Org/Qwen-Image-2.1 VAE
  3. 3.Place files into: ComfyUI/models/vae/
  1. 1.Diffusion Model (This Repo):
  2. 2.Download your preferred .gguf file from above.
  3. 3.Place into: ComfyUI/models/diffusion_models/ (or ComfyUI/models/unet/)

🎨 Included Ready-to-Use Workflows

Both Text-to-Image and Image-to-Image editing workflows are packaged in this repo:

How to Use:

  1. 1.Ensure you have installed [ComfyUI-GGUF](https://github.com/city96/ComfyUI-GGUF) (search for ComfyUI-GGUF inside the ComfyUI Manager).
  2. 2.Drag and drop either of the .json workflow files into your ComfyUI workspace.
  3. 3.In the Unet Loader (GGUF) node, select the downloaded .gguf file.
  4. 4.Verify your CLIP/Text Encoder and VAE node loaders point to the files downloaded above.
  5. 5.Queue prompt and generate!

👏 Acknowledgments & Credits