GGUF, Safetensors, and MLX quantized formats of Qwen-Image-2.1 for local image generation. Try it live, explore quantization options, and learn how to run it with ComfyUI.
Generate images right here in your browser using the Hugging Face Inference API. Enter a prompt and your HF token to get started.
Your generated image will appear here
All quantization formats available in the repository. Q4_K_M is recommended for the best balance of size and quality.
| Quantization | File | Size | Format |
|---|---|---|---|
| BF16 | qwen-image-2.1-UC-BF16.gguf |
14.23 GB | GGUF |
| FP8 | qwen-image-2.1-UC-fp8.safetensors |
6.63 GB | Safetensors |
| INT8 ConvRot | qwen-image-2.1-UC-int8_convrot.safetensors |
6.76 GB | Safetensors |
| NVFP4 | qwen-image-2.1-UC-NVFP4.safetensors |
4.20 GB | Safetensors |
| MLX 4-bit | qwen-image-2.1-UC-MLX-4bit.safetensors |
4.00 GB | MLX |
| MLX 6-bit | qwen-image-2.1-UC-MLX-6bit.safetensors |
5.78 GB | MLX |
| MLX 8-bit | qwen-image-2.1-UC-MLX-8bit.safetensors |
7.56 GB | MLX |
| Q8_0 | qwen-image-2.1-UC-Q8_0.gguf |
7.59 GB | GGUF |
| Q6_K | qwen-image-2.1-UC-Q6_K.gguf |
5.88 GB | GGUF |
| Q5_K_M | qwen-image-2.1-UC-Q5_K_M.gguf |
5.22 GB | GGUF |
| Q4_K_M Recommended | qwen-image-2.1-UC-Q4_K_M.gguf |
4.60 GB | GGUF |
| Q4_0 | qwen-image-2.1-UC-Q4_0.gguf |
4.15 GB | GGUF |
Text encoder and VAE files packaged for ComfyUI.
| Type | File | Precision | Size |
|---|---|---|---|
| Text Encoder | text_encoders/qwen3vl_8b_bf16.safetensors |
BF16 | 17.53 GB |
| Text Encoder | text_encoders/qwen3vl_8b_int8_convrot.safetensors |
Int8 | 9.35 GB |
| VAE | BF16 | 676 MB |
Step-by-step setup to run the uncensored quantized model on your own machine.
Download the files and place them in their respective ComfyUI directories:
ComfyUI/
└── models/
├── diffusion_models/
│ └── qwen-image-2.1-UC-Q4_K_M.gguf
├── text_encoders/
│ └── qwen3vl_8b_bf16.safetensors
└── vae/
└── qwen_image_2.1_vae_bf16.safetensors
Clone the GGUF loader extension for ComfyUI:
cd ComfyUI/custom_nodes
git clone https://github.com/leejet/ComfyUI-GGUF
Use the Unet Loader (GGUF) node to load the .gguf file. For .safetensors files, use the standard CLIPLoader with type qwen_image. Load the VAE with VAELoader pointing to qwen_image_2.1_vae_bf16.safetensors.
For lower VRAM setups, use qwen3vl_8b_int8_convrot.safetensors as the text encoder instead of the BF16 version. You can also launch ComfyUI with --lowvram to reduce memory usage.