Overview
- Tool name: Qwen-Image 3.0 Pro (model ID `qwen-image-3.0-pro`)
- Developer: Qwen team
- Official model page: https://docs.bailian.console.aliyun.com/zh/model-studio/qwen-image-3-0-pro
- Official pricing: Alibaba Cloud Model Studio model pricing page
- Category: AI image models
- Released: August 2026
Qwen-Image 3.0 Pro is the flagship tier of the Qwen image line. Qwen's own framing is blunt: the goal is not merely images that look good, but images that are *usable* - making image generation a tool that actually ships in production. The targets are specific: complex layouts, accurate text, commercial-grade quality.
The standard tier `qwen-image-3.0` sits below it, aimed at high-volume everyday generation at a lower price.
Official pricing (Alibaba Cloud Model Studio)
International (USD)
| Model | Input | 1K output | 2K output | Free quota |
|---|---|---|---|---|
| qwen-image-3.0-pro | $0.003 / image | $0.04 / image | $0.075 / image | 10 images |
| qwen-image-3.0 | $0.003 / image | $0.03 / image | $0.03 / image | 10 images |
Domestic (CNY, China North 2 Beijing)
| Model | Image input | 1K generation | 2K generation |
|---|---|---|---|
| qwen-image-3.0-pro | CNY 0.02 / image | CNY 0.25 / image | CNY 0.5 / image |
Free quota is valid for 90 days from the later of Model Studio activation, model release, or application approval. Pricing verified against the official Model Studio pricing page (October 2026).
Important: international and domestic prices differ, and unit prices vary by region (Singapore, Frankfurt, Virginia, Tokyo, Hong Kong) - check the console figures for the region you actually call.
Specifications (official model page)
- Input modalities: image and text; output modality: image
- Rate limit: RPM 5 (requests per minute), consistent across regions
- Explicitly unsupported: function calling, structured output, web search, prefix continuation, context caching, batch inference, model tuning
That last line matters: if your workflow depends on batch inference or context caching to hold costs down, Qwen-Image 3.0 Pro does not offer them today - you build that in your own scheduling layer.
Relationship to Qwen-Image 2.1
These two lines serve different worlds; do not confuse them:
- Qwen-Image 2.1 (September 20, 2026): 7B open weights, native RGBA transparency, 10 reference images, locally deployable - but under the Qwen Research License, non-commercial only
- Qwen-Image 3.0 Pro: closed commercial API on Alibaba Cloud Model Studio, commercial use allowed, built for complex layouts and accurate text
Choose 3.0 Pro for commercial use, Chinese typography and a hosted service; choose 2.1 for transparency, local deployment and research.
Who it is for
- Chinese-language commercial assets: complex layouts with accurate Chinese and English text - the headline capability
- E-commerce and marketing: product imagery, campaign key visuals, ad variants in bulk
- Domestic settlement and compliance: billed in CNY through Alibaba Cloud Model Studio with local invoicing
- Cost-sensitive high-volume generation: the standard `qwen-image-3.0` at $0.03 per image for both 1K and 2K
- Teams already on Alibaba Cloud: one account, one monitoring stack
Limitations
- Closed source, no weights, no self-hosting
- No batch inference and no context caching, limiting cost optimization at high throughput
- RPM of only 5 - weak concurrency, so expect queuing or a limit increase request
- International, domestic and per-region prices differ, which is easy to misread
- No published head-to-head benchmark against GPT Image, Nano Banana and comparable models
Frequently asked questions
How much is Qwen-Image 3.0 Pro per image?
International: $0.003 per input image, $0.04 at 1K and $0.075 at 2K. Domestic (China North 2 Beijing): CNY 0.02 per input image, CNY 0.25 at 1K and CNY 0.5 at 2K. Regional unit prices vary - trust the Model Studio console.
Is there a free quota?
Yes, 10 images, valid for 90 days from the later of Model Studio activation, model release, or application approval.
How is it different from Qwen-Image 2.1?
2.1 is 7B open weights with native RGBA transparency and local deployment, but research-licensed and non-commercial only. 3.0 Pro is a closed commercial API on Model Studio built for complex layouts and accurate text. Pick 3.0 Pro for commercial work.
Can I call it in batch?
The official model page explicitly marks batch inference as unsupported, with a rate limit of RPM 5. High-throughput needs must be handled in your own scheduler.
Does it support web search?
No. The official capability table lists web search, context caching, prefix continuation, structured output, function calling and model tuning as unsupported.
*This page is not affiliated with Qwen. Specs and pricing verified as of October 2026 (Alibaba Cloud Model Studio official model and pricing pages); confirm current figures in the console.*