Image generation AI 'Qwen-Image-2.0' has been released, and it is strong in live photos and text, and is also integrated with the Edit model.
Qwen (Tongyi Lab), Alibaba's AI research team, announced the image generation AI ' Qwen-Image-2.0 ' on February 10, 2026. Qwen-Image-2.0 is a model that integrates Qwen-Image and Qwen-Image-Edit, and excels at generating realistic images and images that include text.
Qwen-Image-2.0: Professional infographics, exquisite photorealism
Qwen-Image-2.0 has improved the quality of generated images compared to the existing Qwen-Image series. In the AI Arena, a blind AI testing service run by Alibaba, it outperformed Gemini-2.5-Flash-Image-Preview (Nano Banana) in the task of generating images from text.
In the image editing task, it also outperformed Gemini-2.5-Flash-Image-Preview (Nano Banana) and recorded a score comparable to Gemini-3-Pro-Image-Preview (Nano Banana Pro).
Qwen-Image-2.0 boasts a smaller model size than the existing Qwen-Image series, enabling faster inference. It also supports input of 1,000 tokens and allows users to generate slide images while specifying the placement of elements in detail.
Below is the result of creating a 'slide explaining the evolution of the Qwen-Image series so far' using Qwen-Image-2.0. Even complex configurations that include multiple images within an image can be processed without failure.
A realistic image of a forest.
It can also generate images containing long text.
Generate posters by specifying the placement of people in detail.
It can also generate information-dense slides.
Furthermore, it is possible to generate manga while specifying the dialogue and composition for each frame.
Editing tasks are also possible. In the example below, one portrait is input and nine images with different poses are generated.
You can also combine subjects that appear in separate photos into one photo.
Qwen-Image-2.0 can be used on Qwen Chat. The previous Qwen-Image series was also released as an open model that anyone could download, but at the time of writing, the model data for Qwen-Image-2.0 was not publicly available.
Related Posts:
in AI, Posted by log1o_hf