Qwen AI Releases Text-to-Image Model Qwen-Image-3.0: Enhanced Capability in Understanding Lengthy Text, Complex Formatting, and Small Fonts

According to DynaInsight monitoring, Qianwen has released the third-generation image generation model Qwen-Image-3.0. The new model supports a maximum of 4.5k Tokens input, focusing on improving the ability to generate complex layouts, small text, and knowledge-based images.

It can generate newspapers, test papers, short drama storyboards, and 3x3 information grids with a long command. The model can handle multiple themes in the same image, generating formulas, charts, figures, and Chinese and English text simultaneously. An official demonstration of a 3x3 grid contains nine independent content categories, with the command description approximately 3.7k Tokens long.

Text rendering has also been further enhanced. Qianwen stated that the model can clearly generate text as small as 10px and process LaTeX formulas, superscripts, subscripts, handwritten annotations, and dense newspaper layouts. Details such as pores, hair strands, paper texture, and painting textures are also more realistic.

Qwen-Image-3.0 natively supports 12 languages, over 100 artistic styles, and multiple UI interfaces. The officials also demonstrated use cases such as generating weather maps, restoring traditional paintings, and creating professional information graphics online. Compared to merely generating aesthetically pleasing images, this generation emphasizes practical scenarios such as presentations, educational materials, design, and content creation.