base_model, base_model_relation, datasets, frameworks, language, license, tags, tasks

base_model	base_model_relation	datasets	frameworks	language	license	tags	tasks
Qwen/Qwen-Image	quantized	mit-han-lab/svdquant-datasets	PyTorch	en	Apache License 2.0	text-to-image SVDQuant Qwen-Image Diffusion Quantization ICLR2025	text-to-image-synthesis

Model Card for nunchaku-qwen-image

This repository contains Nunchaku-quantized versions of Qwen-Image, designed to generate high-quality images from text prompts, advances in complex text rendering. It is optimized for efficient inference while maintaining minimal loss in performance.

News

• [2025-08-27] 🔥 Release 4-bit 4/8-step lightning Qwen-Image!
• [2025-08-15] 🚀 Release 4-bit SVDQuant quantized Qwen-Image model with rank 32 and 128!

Model Details

Model Description

• Developed by: Nunchaku Team
• Model type: text-to-image
• License: apache-2.0
• Quantized from model: Qwen-Image

Model Files

Data Type: INT4 for non-Blackwell GPUs (pre-50-series), NVFP4 for Blackwell GPUs (50-series). Rank: r32 for faster inference, r128 for better quality but slower inference.

Base Models

Standard inference speed models for general use

Data Type	Rank	Model Name	Comment
INT4	r32	svdq-int4_r32-qwen-image.safetensors
	r128	svdq-int4_r128-qwen-image.safetensors
NVFP4	r32	svdq-fp4_r32-qwen-image.safetensors
	r128	svdq-fp4_r128-qwen-image.safetensors

4-Step Distilled Models

4-step distilled models fused with Qwen-Image-Lightning-4steps-V1.0 LoRA using LoRA strength = 1.0

Data Type	Rank	Model Name	Comment
INT4	r32	svdq-int4_r32-qwen-image-lightningv1.0-4steps.safetensors	Fused with Qwen-Image-Lightning-4steps-V1.0 LoRA
	r128	svdq-int4_r128-qwen-image-lightningv1.0-4steps.safetensors	Fused with Qwen-Image-Lightning-4steps-V1.0 LoRA. Better quality, slower inference
NVFP4	r32	svdq-fp4_r32-qwen-image-lightningv1.0-4steps.safetensors	Fused with Qwen-Image-Lightning-4steps-V1.0 LoRA
	r128	svdq-fp4_r128-qwen-image-lightningv1.0-4steps.safetensors	Fused with Qwen-Image-Lightning-4steps-V1.0 LoRA. Better quality, slower inference

8-Step Distilled Models

8-step distilled models fused with Qwen-Image-Lightning-8steps-V1.1 LoRA using LoRA strength = 1.0

Data Type	Rank	Model Name	Comment
INT4	r32	svdq-int4_r32-qwen-image-lightningv1.1-8steps.safetensors	Fused with Qwen-Image-Lightning-8steps-V1.1 LoRA
	r128	svdq-int4_r128-qwen-image-lightningv1.1-8steps.safetensors	Fused with Qwen-Image-Lightning-8steps-V1.1 LoRA. Better quality, slower inference
NVFP4	r32	svdq-fp4_r32-qwen-image-lightningv1.1-8steps.safetensors	Fused with Qwen-Image-Lightning-8steps-V1.1 LoRA
	r128	svdq-fp4_r128-qwen-image-lightningv1.1-8steps.safetensors	Fused with Qwen-Image-Lightning-8steps-V1.1 LoRA. Better quality, slower inference

Model Sources

• Inference Engine: nunchaku
• Quantization Library: deepcompressor
• Paper: SVDQuant: Absorbing Outliers by Low-Rank Components for 4-Bit Diffusion Models
• Demo: demo.nunchaku.tech

Usage

• Diffusers Usage: See qwen-image.py and qwen-image-lightning.py.
• ComfyUI Usage: See nunchaku-qwen-image.json.

Performance

Citation

@inproceedings{
  li2024svdquant,
  title={SVDQuant: Absorbing Outliers by Low-Rank Components for 4-Bit Diffusion Models},
  author={Li*, Muyang and Lin*, Yujun and Zhang*, Zhekai and Cai, Tianle and Li, Xiuyu and Guo, Junxian and Xie, Enze and Meng, Chenlin and Zhu, Jun-Yan and Han, Song},
  booktitle={The Thirteenth International Conference on Learning Representations},
  year={2025}
}