Qwen-Image is an open-source image generation foundation model from Alibaba’s Qwen team, notable for complex text rendering and precise image editing. Released on GitHub under the Apache-2.0 license with weights on Hugging Face and ModelScope, it has earned more than 8,400 stars, with the repository updated into 2026 alongside newer edit-model releases.
Qwen-Image is a powerful image generation foundation model capable of complex text rendering and precise image editing. Text rendering is its headline strength: the model can lay out multi-line, correctly spelled text inside posters, slides, and product shots, historically a weak point for diffusion models, and it handles both English and Chinese typography. The companion Qwen-Image-Edit models extend the base with instruction-based editing, letting users change objects, styles, and layouts in an existing image while keeping the rest intact. The release includes a technical report, an official blog write-up, and hosted demos for both generation and editing on Hugging Face Spaces, so anyone can try it before downloading weights. Inference code in this repository integrates with the Diffusers ecosystem.
Qwen-Image is free and open source: the code and model weights are released under Apache-2.0, and the online demos are free to try. Self-hosting costs only your GPU compute. The model is also accessible through Qwen’s own chat products, but the open release carries no license fee.
Qwen-Image is for developers and creators who need generated images with accurate embedded text: marketing and e-commerce visuals, poster and slide design, and apps that edit user photos by instruction. ML teams also use it as a strong open baseline for benchmarking text-to-image and editing systems.