Qwen Image Edit GGUF
GGUF conversions of Qwen Image Edit (20B) optimized for local inference.
These models are converted from the official Qwen Image Edit release into the GGUF format while preserving the original architecture and metadata. They are intended for use with ComfyUI GGUF Loader and other compatible GGUF runtimes.
Available Models
| File | Quantization | Size | Recommended VRAM | Description |
|---|---|---|---|---|
| qwen-image-edit-nsfw-abliterated-bf16.gguf | BF16 | 40.9 GB | 48 GB+ | Original precision with maximum image quality |
| qwen-image-edit-nsfw-abliterated-q6_k.gguf | Q6_K | ~16.8 GB | 24 GB+ | Excellent quality with significantly reduced VRAM usage |
| qwen-image-edit-nsfw-abliterated-q5_k.gguf | Q5_K | ~14.3 GB | 20–24 GB | Great balance between quality and model size |
| qwen-image-edit-nsfw-abliterated-q4_k.gguf | Q4_K | 11.5 GB | 16–24 GB | Best balance between image quality, VRAM usage, and storage |
Quantization Guide
BF16
- Original precision
- Highest possible image quality
- No quantization loss
- Recommended for evaluation and high-end GPUs
Q6_K
- 6-bit GGML K-Quant
- Near-BF16 image quality
- Lower VRAM requirements
- Excellent choice for 24 GB GPUs
Q5_K
- 5-bit GGML K-Quant
- Excellent quality-to-size ratio
- Lower VRAM usage than Q6_K
- Recommended for most users with 20–24 GB GPUs
Q4_K
- 4-bit GGML K-Quant
- Approximately 72% smaller than BF16
- Significantly lower VRAM requirements
- Excellent quality-to-size ratio
- Recommended for most users
Features
- ✅ Official Qwen Image Edit converted to GGUF
- ✅ Original metadata preserved
- ✅ Native GGUF format
- ✅ Dynamic VRAM loading support
- ✅ Compatible with ComfyUI GGUF Loader
- ✅ Optimized for local inference
Architecture:
qwen_image
Compatibility
Successfully tested with:
- ComfyUI
- ComfyUI GGUF Loader
- NVIDIA CUDA GPUs
- Dynamic VRAM Loading
Installation
Download your preferred quantization and place it inside:
ComfyUI/models/unet/
Load the model using the ComfyUI GGUF Loader node.
Upcoming Quantizations
Additional quantizations are currently being validated and will be released after testing.
Planned releases:
- Q8_0
Source Model
Official Qwen Image Edit repository:
https://huggingface.co/Qwen/Qwen-Image-Edit
Disclaimer
This repository provides community-maintained GGUF conversions of the official Qwen Image Edit model.
The original model architecture, training, and weights were created by the Qwen Team.
No model weights have been modified beyond format conversion and quantization.
Acknowledgements
- Qwen Team — Original Qwen Image Edit model
- ggufy — GGUF conversion tool
- ComfyUI
- ComfyUI GGUF Loader
License
This repository distributes GGUF conversions of the official Qwen Image Edit model.
The converted weights are released under the same Apache-2.0 license as the original model.
For the complete license terms, please refer to the official repository:
- Downloads last month
- 585
6-bit
16-bit
Model tree for badgame/qwen-image-edit-gguf
Base model
Qwen/Qwen-Image-Edit