Qwen-Image-2.1 Released with Enhanced Image Generation Features

• •Qwen-Image-2.1 integrates text-to-image generation and editing in one model.
• •It supports generating transparent images and editing with up to 10 reference images.
• •The model features a compact architecture with 7 billion parameters for efficient performance.

On September 20, 2026, Qwen.ai announced the open-source release of Qwen-Image-2.1, an advanced image model that combines text-to-image generation and image editing capabilities. The model features a compact architecture with 7 billion parameters and supports generating both regular and transparent images. Key improvements include enhanced inference efficiency, versatile editing options with up to 10 reference images, and realistic textures. The model also integrates native transparency, allowing users to edit transparent layers and extract subjects from photos. This release aims to balance image quality with computational cost, making it suitable for various design tasks. The model is available under the Qwen Research License Agreement.