qwen-image-3.0
qwen-image-3.0 supports text-to-image generation and image editing through synchronous APIs. Each request can return one to six PNG images and use up to 3 ordered reference images. Choose a custom output size or auto, set a seed, and control prompt enhancement and watermarking. OpenAI generations uses the image field to select editing; OpenAI edits and the Bailian native synchronous endpoint are also available. Successful requests return completed image URLs directly. Pricing is in USD and uses the actual successful output count. The price section describes resolution tiers and any reference-image charge. Masks, streaming and Base64 output are unavailable. Balanced speed and quality for image generation and editing.

- Model ID
- qwen-image-3.0
- Provider
- Alibaba
- API endpoint
- /v1/images/generations
- Pricing
- 0.026470588235294118
- Last updated
- 2026-10-27
Capabilities
- Balanced speed and quality for image generation and editing.
- One to six outputs and up to 3 ordered reference images per request.
- Custom output sizes or auto, reproducible seed input, prompt enhancement and watermark controls.
- Synchronous OpenAI generation/editing and Bailian native request formats.
Limitations
- No masks, streaming or Base64 output; completed images are returned as URLs.
- At most 3 reference images; each reference must be at most 10 MB.
- Image generation is probabilistic; a seed does not guarantee identical results.
Use cases
- Product and advertising visuals.
- Reference-based restyling and image revision.
- Programmatic creative workflows with direct synchronous results.
Pricing and billing examples
- Image output | 0.026470588235294118 | output_images
- Reference image input | 0.002941176470588235 | input_images
Request parameters
| Parameter | Type | Required | Default | Options |
|---|---|---|---|---|
| model | string | Yes | ||
| prompt | textarea | Yes | ||
| image | asset_list | No | ||
| size | size | No | 1024x1024 | 1024x1024, 1536x1024, 1024x1536, 1536x1536, 2048x2048, auto |
| n | integer | No | 1 | |
| seed | integer | No | ||
| prompt_extend | boolean | No | true | |
| watermark | boolean | No | false | |
| negative_prompt | textarea | No | ||
| prompt_extend_mode | select | No | direct | direct, agent |
| enable_thinking | boolean | No | true | |
| response_format | select | No | url | url |
API examples
cURL
curl -X POST 'https://api.uniall.ai/v1/images/generations' \
-H 'Authorization: Bearer $UNIALL_API_KEY' \
-H 'Content-Type: application/json' \
-d '{"enable_thinking":true,"model":"qwen-image-3.0","n":1,"prompt":"Describe the output you want","prompt_extend":true,"prompt_extend_mode":"direct","response_format":"url","size":"1024x1024","watermark":false}'Python
import json
import requests
payload = json.loads(r'''{"enable_thinking":true,"model":"qwen-image-3.0","n":1,"prompt":"Describe the output you want","prompt_extend":true,"prompt_extend_mode":"direct","response_format":"url","size":"1024x1024","watermark":false}''')
response = requests.post(
"https://api.uniall.ai/v1/images/generations",
headers={"Authorization": "Bearer " + UNIALL_API_KEY},
json=payload,
)
response.raise_for_status()
result = response.json()JavaScript
const response = await fetch("https://api.uniall.ai/v1/images/generations", {
method: 'POST',
headers: {
Authorization: `Bearer ${process.env.UNIALL_API_KEY}`,
'Content-Type': 'application/json',
},
body: JSON.stringify({"enable_thinking":true,"model":"qwen-image-3.0","n":1,"prompt":"Describe the output you want","prompt_extend":true,"prompt_extend_mode":"direct","response_format":"url","size":"1024x1024","watermark":false}),
});
if (!response.ok) throw new Error(`HTTP ${response.status}`);
const result = await response.json();Frequently asked questions
Which APIs are available?
POST /v1/images/generations, POST /v1/images/edits and POST /api/v1/services/aigc/multimodal-generation/generation use the same API key.
How do I edit an image?
Add image to generations, use JSON or multipart with edits, or put references in native input.messages content. Up to 3 ordered references are supported.
How many images can I request?
Use n=1..6 in OpenAI JSON, or parameters.n=1..6 in native JSON. Final billing counts successfully returned outputs.
How are outputs and references billed?
Output prices are the same across supported sizes. Each reference occurrence is charged once per request. All prices are USD, converted from official Beijing prices at USD 1 = CNY 6.8; group discounts apply.
Which model-specific controls are available?
The 3.0 family supports negative_prompt, prompt_extend_mode and enable_thinking. Agent enhancement supports text-to-image only.
Do I need to poll a task?
No. These endpoints wait for completion and return image URLs directly. Use a suitable client timeout and do not automatically retry an uncertain paid request.