Qwen-Image-2.1

Qwen-Image-2.1 is an open-weight image generation and editing model developed by Alibaba’s Qwen team. Released in September 2026, the model unifies text-to-image creation and instruction-based image editing within a single architecture, allowing prompts and multiple reference images to operate seamlessly in the same workflow.

The model features a lightweight visual generation component comprising 7 billion parameters structured around 32 Single-Stream Diffusion Transformer (DiT) layers. This compact design uses mixed-granularity attention and prefix KV cache reuse to deliver strong visual fidelity and native 2K resolution outputs while maintaining high inference efficiency on consumer-grade hardware.

Key capabilities include native transparency generation, allowing users to output images with a real alpha channel (RGBA) directly from text prompts, alongside multi-reference editing that supports up to 10 input images. The architecture excels at maintaining consistency for people and products, rendering precise typography, and executing local edits using annotations or masks.