Variant	Folder	Size
3-bit	`Qwen-Image-Edit-2511-q3`	~23 GB
4-bit	`Qwen-Image-Edit-2511-q4`	~25 GB
5-bit	`Qwen-Image-Edit-2511-q5`	~28 GB
6-bit	`Qwen-Image-Edit-2511-q6`	~30 GB
8-bit	`Qwen-Image-Edit-2511-q8`	~35 GB

Full (unquantized) weights: ~54 GB from the original repo.

Usage

Download a specific quantization:

huggingface-cli download fcreait/Qwen-Image-Edit-mflux Qwen-Image-Edit-2511-q4 --local-dir ./qwen-edit-q4

Generate with mflux:

bash

mflux-generate-qwen-edit \
  --model ./qwen-edit-q4/Qwen-Image-Edit-2511-q4 \
  --base-model qwen \
  --image-paths input.jpg \
  --prompt "your editing prompt" \
  --output output.png

Details

Quantized with mflux v0.17.4 using mflux-save
Source model: Qwen/Qwen-Image-Edit-2511
Format: MLX safetensors with mflux metadata
Includes VisionTransformer weights (required for image editing)

License

These weights are quantized derivatives of Qwen/Qwen-Image-Edit-2511, released by Alibaba Cloud under the Apache License 2.0. Quantization does not change the license — these derivatives are distributed under the same terms.

Copyright the Qwen team, Alibaba Cloud. Licensed under the Apache License, Version 2.0. A copy of the license is available at http://www.apache.org/licenses/LICENSE-2.0.

Quantization by CREAIT. mflux is MIT licensed.