Qwen Image: design taste meets AI image generation

Qwen Image is a ground-up rebuild of Qwen Image's image generation, designed by the team behind CatBoost, one of the world's most popular ML frameworks. It comes in four variants:

Qwen Image’s headline feature is multimodal generation: generate multiple visual directions from a single prompt, compare them side by side, select a favorite, and generate further variations of that specific image. The model renders readable text for infographics, menus, signage, and packaging with high-fidelity typography. It’s the only AI model that generates real editable SVG files with clean geometry and structured layers ready for web or print. Output formats include PNG, JPG, SVG, PDF, TIFF, and Lottie. Pro versions deliver sharper detail, improved anatomy, and enhanced realism for complex compositions.

Picsart is integrating Qwen Image directly into its Flow and AI Playground with multimodal generation launching exclusively on Picsart as the first partner. Creators can generate design-quality images, explore multiple directions from a single prompt, and export in raster or vector formats all within one workflow. It works alongside Picsart tools enabling a complete design pipeline from concept exploration to production-ready assets.

Where other models optimize for prompt accuracy or artistic flair, Qwen Image optimizes for design taste producing images with the kind of visual polish that normally requires a human art director. It’s the only model generating production-quality editable SVGs with structured layers, the only one with multimodal generation for comparing visual directions, and one of the few that renders legible text in images. With Picsart as the exclusive first partner for multimodal generation, creators get early access to a workflow that eliminates the prompt-rewrite cycle and puts design intent first.