Alibaba releases Qwen-Image-3.0 for image generation with long text and multilingual support

Alibaba released its latest image generation model Qwen-Image-3.0, focusing on practical use and productivity. It handles up to 4.5k token input, creating complex layouts like knowledge diagrams with formulas, geometry, and UI screens in one go. This tackles the issue of models struggling with real-world tasks beyond just making pretty pictures.

The model excels at fine details, rendering text as small as 10 pixels clearly, capturing pores and hair textures, and even restoring damaged ancient paintings. It natively supports 12 languages and over 20 fonts, making text crisp in posters and storyboards, cutting production costs for multilingual materials.

Qwen-Image-3.0 is now available via API on Alibaba Cloud, with free trials coming soon on Qwen Studio and the Qwen app. Alibaba says the goal is to turn image generation into a real productivity tool, not just something that looks good.