Qwen Image

by Alibaba

Qwen Image is a 20B-parameter MMDiT (Multimodal Diffusion Transformer) text-to-image model from Alibaba's Qwen team, released August 2025 under Apache 2.0. Its standout capability is state-of-the-art text rendering - multi-line layouts, paragraphs and fine print - in both alphabetic and logographic scripts (especially Chinese). It pairs a Qwen2.5-VL 7B text encoder with a custom VAE and generates natively at 1328x1328. A refreshed 2512 revision ships in the same Comfy-Org repo.

ImageApache 2.0Released Aug 202520B parameters

Hugging Face repo ↗Alibaba ↗

Drag or click to move along the pipeline

full qualitylightest

Lighter tiers fit smaller GPUs and lose some detail. Start with the heaviest one that loads.

The model

Qwen Image

BF1653.7 GB

Three pieces, three separate downloads.

Diffusion modelGoes in diffusion_models/

Generates the image

38.0 GB↓↗

Text encoderGoes in text_encoders/

Turns your prompt into guidance

15.4 GB↓↗

VAEGoes in vae/

Decodes the result into pixels

0.2 GB↓↗

LoRAs (optional)Goes in loras/

No LoRAs indexed for this model.

SamplerKSampler

Copy these values into your KSampler.

Steps50

CFG4

Samplereuler

Schedulersimple

Native res.1328x1328 (native), up to 1664 long side

Shift3.1

Usage notes

GGUF is the lightest way to run this model.

Apache 2.0. Check the official repo for usage terms.

Click a file name to copy it.