Skip to main content

Qwen Image 3.0 Pro

Generate or edit images at 1K or 2K with the higher quality tier of Qwen Image 3.0.
Qwen Image 3.0 Pro is the higher quality tier of Alibaba’s Qwen Image 3.0. One endpoint covers generation and editing: send a prompt alone, or add up to three images to edit or combine them. It renders at 2K (about 4 megapixels) by default, or at 1K, in 10 aspect ratios.

​
Key capabilities

  • Generation and editing in one endpoint: Text-to-image, or edit with up to 3 images
  • Refer to images by position: image 1, image 2, image 3, in the order you send them
  • Two resolutions: 2k (default, about 4 megapixels) or 1k (about 1 megapixel). It is the only parameter that changes the price
  • Aspect ratios: 10 ratios from 2:1 to 1:2
  • Negative prompt and seed control
  • Prompt range: 1 to 10,000 characters
  • Async processing: Webhook notification or polling for task completion

​
Use cases

  • Product and marketing images: Generate scenes, posters and social graphics from a description
  • Image edits: Change an element, the background or the style of an image while keeping the rest
  • Composition: Combine up to three images, e.g. “put the product from image 2 on the table in image 1”
  • Restyling: Turn a photo into an illustration, watercolor or sketch

​
API operations

Create an image with the POST endpoint, then track it with the list and status endpoints. Generation returns a task ID for async polling or webhook notification.

POST /v1/ai/text-to-image/qwen-image-3-pro

Create a new image generation or editing task

GET /v1/ai/text-to-image/qwen-image-3-pro

List all Qwen Image 3.0 Pro tasks

GET /v1/ai/text-to-image/qwen-image-3-pro/{task-id}

Get task status and results by ID

​
Parameters

​
Output sizes

Shaped to aspect_ratio at the selected resolution. Measured examples:

​
Frequently Asked Questions

Qwen Image 3.0 Pro is Alibaba’s image generation and editing model. You send a prompt (1-10,000 characters), optionally with up to 3 images, and receive a task ID immediately. Poll the task endpoint or configure a webhook_url to receive the image URL when processing completes. Each request generates one image.
Send it in images and describe the change in the prompt, for example “replace the sky in image 1 with a starry night; keep everything else unchanged”. You can add up to two more images and name them by position (image 2, image 3).
Common image formats as a public HTTPS URL, a base64-encoded string or an upload id (upl_img_...). Each image up to 10 MB, ideally 384 to 2048 px per side. Plain http:// URLs are rejected.
The price depends only on resolution: 1k or 2k (the default when you omit it). Input images, aspect ratio and the other parameters do not change it. See the Pricing page for current rates.
They take the same parameters. Qwen Image 3.0 Pro is the higher quality tier and adds resolution (1k or 2k). Qwen Image 3.0 renders at about 1 megapixel at a flat, lower price.
Rate limits depend on your subscription tier. See Rate Limits for current limits.

​
Best practices

  • Name images by position: Refer to inputs as image 1, image 2, image 3 in the order you send them
  • Say what stays: In edits, describe the change and add “keep everything else unchanged”
  • Use negative_prompt to keep unwanted elements out instead of describing them in the prompt
  • Production integration: Use webhook_url instead of polling for scalable applications
  • Error handling: Implement retry logic with exponential backoff for 503 errors