Qwen Image 3 Pro Edit

• 🔧 Instruction-driven image editing and inpainting from Alibaba's Qwen Image line
• 🆕 Pro-tier editing variant of the third-generation Qwen Image family
• 👁️ Takes a source image plus a natural-language edit instruction
• 🎯 Documented edits include object add/remove/replace, style transfer, in-image text changes
• 📏 Localized edits aim to leave untouched regions consistent
• 🏢 Listed in this catalog as the high-fidelity inpainting option with strong text rendering
• 🌐 Available through Alibaba Cloud Model Studio and third-party inference platforms
• 📚 Alibaba's edit docs note newer editing models accept multiple images, return multiple outputs

Alibaba

Alibaba Group is a Chinese multinational technology company founded in 1999 and headquartered in Hangzhou, Zhejiang. Originally built around e-commerce and cloud computing, Alibaba has become one of the most prolific contributors to open-weight AI research,…

Qwen Image 3 Pro Edit is the professional-tier editing model in the third generation of Alibaba's Qwen Image family. Where a text-to-image sibling such as Qwen Image 3 Pro starts from a prompt alone, this variant consumes an existing image together with a natural-language instruction, which makes it suited to inpainting, object insertion or removal, style and appearance changes, and rewriting text that appears inside a picture, per Alibaba's Model Studio editing documentation.

Compared with the previous generation's Pro editor, Qwen Image 2 Pro, the main documented shifts are in interface and control rather than disclosed architecture. Alibaba's editing guide notes that newer models in the Qwen-Image-Edit series accept multiple input images and can return multiple output images in a single call, whereas earlier editing endpoints were limited to a single input and a single result. In this catalog the Pro variant is the high-fidelity inpainting entry, described as having strong text rendering.

Within this catalog it sits alongside the standard-tier Qwen Image 3 Edit and the base generator Qwen Image 3, so a composition can be produced by the text-to-image model and then refined by the editing variants.

No independently verified benchmark results for this specific release were available at the time of writing, so the description above is limited to documented capabilities from Alibaba Cloud Model Studio.