On September 20, Qwen open-sourced Qwen-Image-2.1, unifying text-to-image generation and image editing in a single model. The visual-generation part has 7B parameters and natively supports 2K resolution, RGBA transparent backgrounds and up to 10 reference images, with local editing via brush, scribble and mask. It supports ComfyUI and Diffusers from day one.
For AIGC creators, the upgrade is highly practical: a single model handles generation, inpainting, style transfer and multi-reference composition, removing the need to chain separate tools. The transparent-background and multi-reference capabilities directly target e-commerce, design and content-production workflows.
The release continues Alibaba's open-weight strategy alongside Qwen3.8-Omni-Flash, reinforcing a trend where domestic models compete on openness and downstream tooling as much as on benchmark scores. It also lowers the barrier for small teams to build image pipelines on Chinese foundational models.
Source: Qwen / Weibo AIGC Daily
