Image generation and editing capabilities are unified in the latest open-weights release from the Qwen series. This 7B architecture allows for native 2K resolution and processes up to 10 reference images for precise local control over portraits and products. The model natively produces RGBA layers, enabling seamless compositing and text editing within transparent images, which improves upon the visual quality of its predecessor.

Integration for the model is available via vLLM-Omni and ComfyUI with day-0 support. The vLLM version implements a 7.1B Diffusion Transformer combined with a Qwen3-VL-8B vision model to speed up inference for multi-image prompts. This configuration focuses on aesthetic improvements for text rendering and supports specialized outputs such as infographics and virtual try-ons.