Alibaba Unveils Qwen-image-2.1: Advanced Open-weight Model for Image Creation

WHO:

MarkTechPost - Alibaba's Qwen team is responsible for the release of Qwen-Image-2.1.

WHAT:

MarkTechPost - Alibaba released Qwen-Image-2.1, a 7B diffusion transformer model for text-to-image generation and multi-reference editing.

Techmeme - Qwen-Image-2.1 is claimed to outperform most closed-source models.

MarkTechPost - The model supports native RGBA transparency and can handle up to 10 reference images for edits.

MarkTechPost - The weights of Qwen-Image-2.1 are public, but commercial use requires a separate license.

HOW:

MarkTechPost - A prefix KV cache is utilized to speed up edits when using up to 10 reference images.

(UPDATED September 21, 2026 17:03 UTC)

Alibaba has unveiled Qwen-Image-2.1, a powerful 7B diffusion transformer model designed for text-to-image generation and multi-reference editing. This new model is said to outperform many closed-source alternatives and supports native RGBA transparency, allowing users to edit with up to 10 reference images. While the weights of Qwen-Image-2.1 are publicly available, commercial use will require a separate license, and a prefix KV cache is employed to enhance editing speed.