Official Qwen Image

These models are based on our own deployments on cloud GPUs. Some of the models are open-source, while others have been fine-tuned or developed in-house by us.

Qwen-Image, launched by Alibaba's Tongyi Qianwen team, is an open-source MMDiT-based image generation model featuring 20 billion parameters (20B). As the first image-generation model in the Qwen series, it excels in complex text rendering, supporting mixed Chinese and English, paragraph-level formatting, and automatic layout design—enabling the creation of high-fidelity text and image effects. Additionally, the model boasts precise image-editing capabilities, allowing users to perform tasks such as text modification, object replacement, and style transfer. Qwen-Image is regarded as combining commercial-grade quality with open-source versatility, delivering leading performance across multiple benchmark tests, particularly excelling in board and document-oriented scenarios.