Skip to content
Model infoModelsSnapshot: 15 Oct 2026

Alibaba Cloud / Qwen Image 3.0

Qwen Image 3.0 Pro

Image generation and editing with thinking and prompt enhancement, plus separate input-image and output-image charges.

availableClosed weightsImage generationImage editing
Official documentation (opens in a new tab)Reviewed 14 Oct 2026
37 of 44 specification fields collected
Output resolution512×512 to 2048×2048 total pixelsSee specification notes below
Reference images33
Input billingSee ratesAlibaba Model Studio · Singapore · International
Output billingSee ratesPublished units and conditions below

Specifications

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit
Not collected
Maximum input tokens
Not collected

The API recommends at most 4,500 prompt tokens; it does not establish a hard maximum.

Maximum output
Not collected
Reasoning
SupportedSource for Reasoning (opens in a new tab)

enable_thinking=true by default; requires prompt_extend=true. Supports direct generation/editing and agent-mode text-to-image, not agent-mode editing.

Streaming
Not collected
Knowledge cutoff
Not collected
Release date
2026-07-20Source for Release date (opens in a new tab)

Earliest reviewed regional availability. A date suffix is not substituted for the platform release date.

Parameters
Not collected
Architecture
Not collected
Languages
Chinese, EnglishSource for Languages (opens in a new tab)

Documented prompt languages. Native text-rendering language marketing is a separate claim.

API aliases
qwen-image-3.0-proSource for API aliases (opens in a new tab)

Rolling names and dated snapshots remain separate; capability comparisons do not prove identical private weights.

Images per request
1–6 output images, default 1Source for Images per request (opens in a new tab)
Output resolutions
512×512 to 2048×2048 total pixelsSource for Output resolutions (opens in a new tab)
Input image formats
JPG, JPEG, PNG, BMP, TIFF, WEBP, GIFSource for Input image formats (opens in a new tab)
Output link expiry
Generated image URLs expire after 24 hoursSource for Output link expiry (opens in a new tab)

Not a complete input-data retention or training-use policy.

Random seed range
0–2,147,483,647; omitted selects a random seedSource for Random seed range (opens in a new tab)

Probabilistic generation does not guarantee identical output for the same seed.

Visible watermark
Optional watermark, off by defaultSource for Visible watermark (opens in a new tab)
Recommended prompt tokens
4,500Source for Recommended prompt tokens (opens in a new tab)

API calls this a recommended maximum, not a verified hard context limit. Model-card context limits are undisclosed.

Input image limits
0 images for generation, 1–3 for editing; 10 MB each; 1–8,000 pixels per side (384–2,048 recommended), longest-to-shortest side ratio at most 10:1. Public URL or Base64.Source for Input image limits (opens in a new tab)
Aspect ratios
Output 1:8 to 8:1Source for Aspect ratios (opens in a new tab)
Resolution controls
Auto-selected from the prompt when omitted. DashScope uses width*height; compatible Images API uses widthxheight or auto.Source for Resolution controls (opens in a new tab)
Prompt rewriting
prompt_extend=true by default; direct mode supports generation/editing. Agent mode supports text-to-image only; agent editing is rejected.Source for Prompt rewriting (opens in a new tab)
Negative prompts
Supported for Qwen Image 3.0; no maximum length is specified in the reviewed unified API.Source for Negative prompts (opens in a new tab)
Request mode
Three transports of one model: compatible Images API (synchronous, non-streaming), DashScope synchronous, and DashScope asynchronous task polling.Source for Request mode (opens in a new tab)
Compatible Images API
POST /compatible-mode/v1/images/generations for both generation and editing; image is a vendor extension. No /images/edits, multipart upload, mask, partial images or streaming. response_format=b64_json is ignored; results are image URLs.Source for Compatible Images API (opens in a new tab)
Image billing tiers
Both input and output tariff tiers use final OUTPUT pixel area: 1K when area ≤ 2,250,000; 2K when area > 2,250,000. Input-image resolution does not independently select the tier.Source for Image billing tiers (opens in a new tab)
Output billing
Input image unit price × input count + output image unit price × successful output count. Failed requests are not charged.Source for Output billing (opens in a new tab)
Usage metering
Reference-image input + successful image output; prices depend on output pixel areaSource for Usage metering (opens in a new tab)
Usage metering
input_image_count/output_image_count and qima_input/output_1k/2k meters are image usage, not token usage.Source for Usage metering (opens in a new tab)
Published availability regions
China (Beijing), Singapore, Germany (Frankfurt), US (Virginia), Japan (Tokyo), China (Hong Kong)Source for Published availability regions (opens in a new tab)

Access & pricing

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

Alibaba Model Studio · Singapore · International

Provider source (opens in a new tab)
API model ID
qwen-image-3.0-pro
Access
POST https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation (synchronous). Alternative compatible Images route: https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1/images/generations (synchronous, non-streaming). Async route: https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1/services/aigc/image-generation/generation with X-DashScope-Async: enable; poll /api/v1/tasks/{task_id}. Match workspace, API key and region. Deployment scope: International.
Regions
Singapore
Deployment scope
InternationalSource for Deployment scope (opens in a new tab)

Explicit model-card scope

Published request quota
5 requests per minuteSource for Published request quota (opens in a new tab)
Task submission rate
5 requests per minuteSource for Task submission rate (opens in a new tab)

Central quota table units retained; synchronous concurrency does not bypass submission limits.

Concurrent requests
No synchronous concurrency cap listed; up to 10 asynchronous tasksSource for Concurrent requests (opens in a new tab)

Synchronous interface unlimited / Asynchronous interface 10

Quota scope
Primary account aggregate across RAM users, workspaces and API keys; per-model quotaSource for Quota scope (opens in a new tab)
Request mode
Compatible Images synchronous (non-streaming); DashScope synchronous or asynchronousSource for Request mode (opens in a new tab)
Free usage allowance
10 images; 90 days from latest activation, release or approvalSource for Free usage allowance (opens in a new tab)

The reviewed quota row does not define a separate input/output allocation.

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 14 Oct 2026
CategoryUsageRateUnit
Image generationInput image · 1K output tier1$0.003image
Image generationInput image · 2K output tier1$0.003image
Image generationOutput image · 1K output tier1$0.04image
Image generationOutput image · 2K output tier1$0.075image

Rate conditions

  1. Input image unit price × input count + output image unit price × successful output count. Failed requests are not charged. Both input and output tariff tiers use final OUTPUT pixel area: 1K when area ≤ 2,250,000; 2K when area > 2,250,000. Input-image resolution does not independently select the tier. Native USD rate; International. No text-token or per-question estimate.
Pricing source (opens in a new tab)

Data & enterprise terms

Image links last 24 hours. Input retention, training use and enterprise regional policy are not established by this duration.

Alibaba Model Studio · China (Beijing) · Scope not stated

Provider source (opens in a new tab)
API model ID
qwen-image-3.0-pro
Access
POST https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation (synchronous). Alternative compatible Images route: https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/compatible-mode/v1/images/generations (synchronous, non-streaming). Async route: https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/api/v1/services/aigc/image-generation/generation with X-DashScope-Async: enable; poll /api/v1/tasks/{task_id}. Match workspace, API key and region. Deployment scope: Scope not stated.
Regions
China (Beijing)
Deployment scope
Scope not statedSource for Deployment scope (opens in a new tab)

The regional card and price row do not label a deployment scope.

Published request quota
5 requests per minuteSource for Published request quota (opens in a new tab)
Task submission rate
5 requests per minuteSource for Task submission rate (opens in a new tab)

Central quota table units retained; synchronous concurrency does not bypass submission limits.

Concurrent requests
No synchronous concurrency cap listed; up to 10 asynchronous tasksSource for Concurrent requests (opens in a new tab)

Synchronous interface unlimited / Asynchronous interface 10

Quota scope
Primary account aggregate across RAM users, workspaces and API keys; per-model quotaSource for Quota scope (opens in a new tab)
Request mode
Compatible Images synchronous (non-streaming); DashScope synchronous or asynchronousSource for Request mode (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 14 Oct 2026
CategoryUsageRateUnit
Image generationInput image · 1K output tier1$0.00275image
Image generationInput image · 2K output tier1$0.00275image
Image generationOutput image · 1K output tier1$0.03438image
Image generationOutput image · 2K output tier1$0.068761image

Rate conditions

  1. Input image unit price × input count + output image unit price × successful output count. Failed requests are not charged. Both input and output tariff tiers use final OUTPUT pixel area: 1K when area ≤ 2,250,000; 2K when area > 2,250,000. Input-image resolution does not independently select the tier. Native USD rate; Scope not stated. No text-token or per-question estimate.
Pricing source (opens in a new tab)

Data & enterprise terms

Image links last 24 hours. Input retention, training use and enterprise regional policy are not established by this duration.

Alibaba Model Studio · Germany (Frankfurt) · Global

Provider source (opens in a new tab)
API model ID
qwen-image-3.0-pro
Access
POST https://{WorkspaceId}.eu-central-1.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation (synchronous). Alternative compatible Images route: https://{WorkspaceId}.eu-central-1.maas.aliyuncs.com/compatible-mode/v1/images/generations (synchronous, non-streaming). Async route: https://{WorkspaceId}.eu-central-1.maas.aliyuncs.com/api/v1/services/aigc/image-generation/generation with X-DashScope-Async: enable; poll /api/v1/tasks/{task_id}. Match workspace, API key and region. Deployment scope: Global.
Regions
Germany (Frankfurt)
Deployment scope
GlobalSource for Deployment scope (opens in a new tab)

Explicit model-card scope

Published request quota
5 requests per minuteSource for Published request quota (opens in a new tab)
Task submission rate
5 requests per minuteSource for Task submission rate (opens in a new tab)

Central quota table units retained; synchronous concurrency does not bypass submission limits.

Concurrent requests
No synchronous concurrency cap listed; up to 10 asynchronous tasksSource for Concurrent requests (opens in a new tab)

Synchronous interface: no limit / Asynchronous interface: 10

Quota scope
Primary account aggregate across RAM users, workspaces and API keys; per-model quotaSource for Quota scope (opens in a new tab)
Request mode
Compatible Images synchronous (non-streaming); DashScope synchronous or asynchronousSource for Request mode (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 14 Oct 2026
CategoryUsageRateUnit
Image generationInput image · 1K output tier1$0.00275image
Image generationInput image · 2K output tier1$0.00275image
Image generationOutput image · 1K output tier1$0.03438image
Image generationOutput image · 2K output tier1$0.068761image

Rate conditions

  1. Input image unit price × input count + output image unit price × successful output count. Failed requests are not charged. Both input and output tariff tiers use final OUTPUT pixel area: 1K when area ≤ 2,250,000; 2K when area > 2,250,000. Input-image resolution does not independently select the tier. Native USD rate; Global. No text-token or per-question estimate.
Pricing source (opens in a new tab)

Data & enterprise terms

Image links last 24 hours. Input retention, training use and enterprise regional policy are not established by this duration.

Alibaba Model Studio · US (Virginia) · Global

Provider source (opens in a new tab)
API model ID
qwen-image-3.0-pro
Access
POST https://{WorkspaceId}.us-east-1.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation (synchronous). Alternative compatible Images route: https://{WorkspaceId}.us-east-1.maas.aliyuncs.com/compatible-mode/v1/images/generations (synchronous, non-streaming). Async route: https://{WorkspaceId}.us-east-1.maas.aliyuncs.com/api/v1/services/aigc/image-generation/generation with X-DashScope-Async: enable; poll /api/v1/tasks/{task_id}. Match workspace, API key and region. Deployment scope: Global.
Regions
US (Virginia)
Deployment scope
GlobalSource for Deployment scope (opens in a new tab)

Explicit model-card scope

Published request quota
5 requests per minuteSource for Published request quota (opens in a new tab)
Task submission rate
5 requests per minuteSource for Task submission rate (opens in a new tab)

Central quota table units retained; synchronous concurrency does not bypass submission limits.

Concurrent requests
No synchronous concurrency cap listed; up to 10 asynchronous tasksSource for Concurrent requests (opens in a new tab)

Synchronous interface unlimited / Asynchronous interface 10

Quota scope
Primary account aggregate across RAM users, workspaces and API keys; per-model quotaSource for Quota scope (opens in a new tab)
Request mode
Compatible Images synchronous (non-streaming); DashScope synchronous or asynchronousSource for Request mode (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 14 Oct 2026
CategoryUsageRateUnit
Image generationInput image · 1K output tier1$0.00275image
Image generationInput image · 2K output tier1$0.00275image
Image generationOutput image · 1K output tier1$0.03438image
Image generationOutput image · 2K output tier1$0.068761image

Rate conditions

  1. Input image unit price × input count + output image unit price × successful output count. Failed requests are not charged. Both input and output tariff tiers use final OUTPUT pixel area: 1K when area ≤ 2,250,000; 2K when area > 2,250,000. Input-image resolution does not independently select the tier. Native USD rate; Global. No text-token or per-question estimate.
Pricing source (opens in a new tab)

Data & enterprise terms

Image links last 24 hours. Input retention, training use and enterprise regional policy are not established by this duration.

Alibaba Model Studio · Japan (Tokyo) · Global

Provider source (opens in a new tab)
API model ID
qwen-image-3.0-pro
Access
POST https://{WorkspaceId}.ap-northeast-1.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation (synchronous). Alternative compatible Images route: https://{WorkspaceId}.ap-northeast-1.maas.aliyuncs.com/compatible-mode/v1/images/generations (synchronous, non-streaming). Async route: https://{WorkspaceId}.ap-northeast-1.maas.aliyuncs.com/api/v1/services/aigc/image-generation/generation with X-DashScope-Async: enable; poll /api/v1/tasks/{task_id}. Match workspace, API key and region. Deployment scope: Global.
Regions
Japan (Tokyo)
Deployment scope
GlobalSource for Deployment scope (opens in a new tab)

Explicit model-card scope

Published request quota
5 requests per minuteSource for Published request quota (opens in a new tab)
Task submission rate
5 requests per minuteSource for Task submission rate (opens in a new tab)

Central quota table units retained; synchronous concurrency does not bypass submission limits.

Concurrent requests
No synchronous concurrency cap listed; up to 10 asynchronous tasksSource for Concurrent requests (opens in a new tab)

Synchronous interface unlimited / Asynchronous interface 10

Quota scope
Primary account aggregate across RAM users, workspaces and API keys; per-model quotaSource for Quota scope (opens in a new tab)
Request mode
Compatible Images synchronous (non-streaming); DashScope synchronous or asynchronousSource for Request mode (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 14 Oct 2026
CategoryUsageRateUnit
Image generationInput image · 1K output tier1$0.00275image
Image generationInput image · 2K output tier1$0.00275image
Image generationOutput image · 1K output tier1$0.03438image
Image generationOutput image · 2K output tier1$0.068761image

Rate conditions

  1. Input image unit price × input count + output image unit price × successful output count. Failed requests are not charged. Both input and output tariff tiers use final OUTPUT pixel area: 1K when area ≤ 2,250,000; 2K when area > 2,250,000. Input-image resolution does not independently select the tier. Native USD rate; Global. No text-token or per-question estimate.
Pricing source (opens in a new tab)

Data & enterprise terms

Image links last 24 hours. Input retention, training use and enterprise regional policy are not established by this duration.

Alibaba Model Studio · China (Hong Kong) · Global

Provider source (opens in a new tab)
API model ID
qwen-image-3.0-pro
Access
POST https://{WorkspaceId}.cn-hongkong.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation (synchronous). Alternative compatible Images route: https://{WorkspaceId}.cn-hongkong.maas.aliyuncs.com/compatible-mode/v1/images/generations (synchronous, non-streaming). Async route: https://{WorkspaceId}.cn-hongkong.maas.aliyuncs.com/api/v1/services/aigc/image-generation/generation with X-DashScope-Async: enable; poll /api/v1/tasks/{task_id}. Match workspace, API key and region. Deployment scope: Global.
Regions
China (Hong Kong)
Deployment scope
GlobalSource for Deployment scope (opens in a new tab)

Explicit model-card scope

Published request quota
5 requests per minuteSource for Published request quota (opens in a new tab)
Task submission rate
5 requests per minuteSource for Task submission rate (opens in a new tab)

Central quota table units retained; synchronous concurrency does not bypass submission limits.

Concurrent requests
No synchronous concurrency cap listed; up to 10 asynchronous tasksSource for Concurrent requests (opens in a new tab)

Synchronous interface unlimited / Asynchronous interface 10

Quota scope
Primary account aggregate across RAM users, workspaces and API keys; per-model quotaSource for Quota scope (opens in a new tab)
Request mode
Compatible Images synchronous (non-streaming); DashScope synchronous or asynchronousSource for Request mode (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 14 Oct 2026
CategoryUsageRateUnit
Image generationInput image · 1K output tier1$0.00275image
Image generationInput image · 2K output tier1$0.00275image
Image generationOutput image · 1K output tier1$0.03438image
Image generationOutput image · 2K output tier1$0.068761image

Rate conditions

  1. Input image unit price × input count + output image unit price × successful output count. Failed requests are not charged. Both input and output tariff tiers use final OUTPUT pixel area: 1K when area ≤ 2,250,000; 2K when area > 2,250,000. Input-image resolution does not independently select the tier. Native USD rate; Global. No text-token or per-question estimate.
Pricing source (opens in a new tab)

Data & enterprise terms

Image links last 24 hours. Input retention, training use and enterprise regional policy are not established by this duration.

Caching

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

01 / INITIAL INPUTProcess & retain

The system processes eligible input and retains reusable state.

02 / LATER REQUESTCheck available state

Reuse depends on matching input, session state and the cache policy.

03 / CACHE HITReuse & generate

Reusing state can reduce repeated work. Any charges follow the billing rules below.

What the measurements say

No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

The record behind the model

Why it is included

Classification inference: the separately documented Qwen Image 2.0/3.0 API selector has no exact checkpoint in the reviewed official Qwen image-weight inventory. Public Qwen-Image, Edit, Layered and 2.1 checkpoints are excluded; private lineage and parameter counts are not inferred from them.

Classification evidence (opens in a new tab)

Lifecycle dates

Platform release · Singapore20 Jul 2026Source (opens in a new tab)
Platform release · China (Beijing)20 Jul 2026Source (opens in a new tab)
Platform release · Hong Kong (China)6 Aug 2026Source (opens in a new tab)
Platform release · Germany (Frankfurt)6 Aug 2026Source (opens in a new tab)
Platform release · Japan (Tokyo)6 Aug 2026Source (opens in a new tab)

Sources reviewed