Qwen-Image-2.1 7B model fuses generation, editing, and transparency

Qwen-Image-2.1 is a 7B single-stream DiT open model combining text-to-image generation, high-fidelity editing (up to 10 reference images), and native RGBA transparency via a 64-channel autoencoder. It offers block-causal attention and prefix KV caching for efficient inference, with vLLM deployment options.