Qwen-Image-2.1-Pro

Qwen-Image-2.1-Pro is a high-tier image generation and editing model developed by Alibaba as part of the Qwen series. Built upon a 7-billion-parameter visual generation component featuring 32 Single-Stream Diffusion Transformer layers, the model unifies text-to-image creation and complex image manipulation within a single architecture. It is engineered to balance high generation fidelity, inference efficiency, and advanced functionality for professional use cases.

Key Capabilities

The model introduces robust features tailored for advanced graphic design and visual production. It natively supports transparent image generation and editing (RGBA), allowing users to create or modify transparent layers and extract subjects seamlessly. Additionally, multi-reference editing enables the incorporation of up to 10 reference images to preserve personal or product identities and combine assets into coherent compositions. Localized edits can be directed through manual sketches, circles, or masks.

Architecture and Performance

Optimized for detailed visual output, Qwen-Image-2.1-Pro provides fine text rendering, accurate portrait lighting, and realistic textures. As part of the higher-quality tier in the Qwen-Image-2.1 ecosystem, the Pro variant prioritizes output fidelity and advanced rendering quality over raw speed, complementing faster checkpoints.