Alibaba releases Qwen-Image-2.1, a 7B open-weight image model
The Decoder reports support for transparent images and up to 10 reference images on powerful consumer GPUs. The research license bars commercial use, which requires a separate Qwen license.
TLDR
The Decoder reports that Qwen-Image-2.1 generates and edits images, and highlights claims that the 7-billion-parameter model beats closed models in image generation. Commercial use requires a separate Qwen license rather than the research license. One early user says they made an image using a prompt specifying messy hair, a collarbone tattoo and a lacy blouse. They also say that prompt was restricted on GPT Image 2.5 and Muse.
Alibaba releases Qwen-Image-2.1, a 7B open-weight image model
The Decoder reports support for transparent images and up to 10 reference images on powerful consumer GPUs. The research license bars commercial use, which requires a separate Qwen license.