Qwen-Image-2.1: 7B Model Combines Generation, Editing, and Transparency

Alibaba's Qwen-Image-2.1 is a 7B-parameter open-source model that unifies text-to-image generation, editing, and native transparent channel (RGBA) output. It supports up to 10 reference images for high-fidelity editing, uses block-causal attention with prefix KV cache for efficiency, and includes a 64-channel RGBA autoencoder for native transparency. The model is optimized for deployment with vLLM and supports 8 chip platforms.