Qwen Open-Sources Qwen-Image-2.1, a 7B Unified Image Generation and Editing Model with Transparent Image Support

The Qwen team has open-sourced Qwen-Image-2.1, unifying text-to-image generation and image editing into a single model. Its visual generation component contains only 7B parameters and natively supports generating and editing transparent images. The model supports up to 10 reference images, enabling localized edits through circles, scribbles, or individual masks. Architecturally, the model adopts mixed-granularity attention and reuses KV caches to enhance inference efficiency, while also improving text rendering, portrait lighting, and subject/product fidelity across tasks such as panoramic imagery, infographics, and storyboarding.