Qwen Image 2.1 Pro
Qwen Image 2.1 Pro is Alibaba's flagship unified model for high-fidelity text-to-image synthesis and multi-modal image editing. Built on an optimized 7B single-stream DiT architecture, it seamlessly merges studio-grade 2K image rendering, multi-image reference control (up to 10 input images), and localized inpainting via mask or brush interactions while strictly preserving subject identity.
| Size | Unit Price (USD/Image) |
|---|---|
1K: 1024*1024, 1280*720, 720*1280, 1024*768, 768*1024 |
0.04 |
2K: 2048*2048, 2560*1440, 1440*2560, 2048*1536, 1536*2048 |
0.04 |
Related Models
Qwen Image 2.1 Turbo
Qwen Image 2.1 Turbo is Alibaba's high-speed, cost-optimized edition of the flagship Qwen Image 2.1 vision model. Engineered on a refined 7B single-stream DiT architecture, it compresses generation down to just 8 denoising steps without sacrificing output fidelity. Qwen Image 2.1 Turbo unifies rapid 2K image rendering, native transparent background (RGBA) generation, and intuitive multi-image reference editing into a single stream.
Qwen Image 3.0 Spicy
Qwen Image 3.0 Spicy is a flagship uncensored image generation model built upon Alibaba's advanced Qwen vision architecture. Operating as a Spicy uncensored model, it completely bypasses standard content filters to unlock maximum creative freedom across photorealistic portraiture, complex visual storytelling, and high-detail stylized concept art without content restrictions. Leveraging enhanced rendering capacities, it delivers sharp textures, realistic lighting physics, and exceptional prompt adherence. Qwen Image 3.0 Spicy is the premier choice for professional designers, digital creators, marketing agencies, and visual artists requiring unconstrained, high-quality image synthesis.
Qwen Image 3.0 Pro Spicy
Qwen Image 3.0 Pro Spicy is a flagship uncensored image generation model built upon Alibaba's advanced Qwen vision architecture. Operating as a Spicy uncensored model, it completely bypasses standard content filters to unlock unrestricted artistic freedom across photorealistic portraiture, complex visual storytelling, and high-detail stylized concept art. Leveraging Pro-tier enhancements, it delivers ultra-sharp textures, master-level lighting physics, fine typography rendering inside images, and exceptional prompt adherence. Qwen Image 3.0 Pro Spicy is the premier choice for professional designers, digital creators, marketing agencies, and visual artists requiring unconstrained, studio-quality image synthesis.
Qwen Image 3.0
Qwen Image 3.0 is Alibaba's next-generation open-weights AI image generation and editing model built on a Diffusion Transformer (DiT) architecture. Supporting both text-to-image (T2I) and image-to-image instruction editing (I2I), it excels at rich content rendering, hyper-realistic detail, and deep contextual understanding. It achieves industry-leading text-in-image typography, rendering clear small text down to 10px across 12 native languages and 20+ fonts for complex multi-column layouts, UI designs, and graphic documents. With open weights available for self-hosting, Qwen Image 3.0 provides highly cost-effective, high-precision visual generation for localized advertising, product UI prototyping, e-commerce graphics, and creative publishing.
Read Me
Qwen Image 2.1 Pro API
Qwen Image 2.1 Pro is a second-generation visual generation foundation model developed by Alibaba's Qwen team. It excels in text-to-image (T2I) generation, image-to-image (I2I) editing, multi-image conditional generation with up to 10 reference images, native RGBA transparent image generation, multilingual text rendering, and native 2K resolution direct output.
On the iCreat platform, this model is accessed via an asynchronous two-step API. The endpoint is https://api.icreat.ai/v1/task/submit/aliyun/qwen-image-2-1-pro, and results are polled via https://api.icreat.ai/v1/task/result. It supports a total pixel range between 512×512 and 2048×2048 with an aspect ratio of 1:8 to 8:1. The pricing is a flat $0.04 per output image for both 1K and 2K tiers, with no charge for input reference images.
Model Positioning
Qwen Image 2.1 Pro is positioned as a versatile and powerful Pro-tier image generation and editing model. It is designed for users who require high flexibility in reference image inputs (up to 10 images) and specialized outputs like native RGBA transparent images. With a flat pricing model of $0.04 per image regardless of 1K or 2K resolution, it is highly cost-effective for complex multi-image conditioning and high-resolution generation tasks.
Core Capabilities
Text-to-Image and Image Editing
The model supports both text-to-image generation and image-to-image instruction-based editing, allowing users to modify existing images or create new ones from textual prompts.
Multi-Image Conditional Generation
It accepts up to 10 reference images, enabling complex multi-image conditioning, style transfer, and continuous editing workflows.
Native RGBA Transparent Output
The model natively supports the generation of RGBA transparent images, which is ideal for design assets, stickers, and layered compositions.
Native 2K Resolution
It can directly output images with a total pixel count up to 2048×2048, delivering high-resolution results without requiring post-processing upscaling.
Flexible Resolution and Auto-Prediction
The total pixel output can be flexibly set between 512×512 and 2048×2048 with an aspect ratio from 1:8 to 8:1. If the size parameter is omitted, the model automatically predicts the optimal resolution and aspect ratio based on the prompt.
Pricing
| Size Tier | Example Resolutions | Unit Price (USD/Image) |
|---|---|---|
| 1K | 10241024, 1280720, 7201280, 1024768, 768*1024 | $0.04 |
| 2K | 20482048, 25601440, 14402560, 20481536, 1536*2048 | $0.04 |
Note: Billed per output image at a flat rate of $0.04 for both 1K and 2K tiers. Input reference images are not charged. The official Alibaba Cloud Bailian pricing is ¥0.25/image (rate-limited to 20 RPM); the iCreat platform price is $0.04/image, subject to the iCreat platform price.
Application Scenarios
- Multi-image style transfer and continuous image editing
- Generation of transparent PNG assets for UI design and e-commerce
- High-resolution 2K poster and illustration generation
- Multilingual text image rendering for marketing materials
- Complex creative workflows requiring multiple visual references
Model Comparison
Comparison Table 1: Qwen Image Series
| Model | Capabilities | Reference Images | Resolution & Pricing | Special Features |
|---|---|---|---|---|
| Qwen Image 2.1 Pro | T2I + I2I/Editing | Up to 10 | 512×512–2048×2048; Flat $0.04/img (1K/2K) | Native RGBA transparent output |
| Qwen Image 3.0 Pro | T2I + I2I/Editing | 1–3 | 1K $0.04/img, 2K $0.075/img | 12-language complex text rendering |
| Qwen Image 3.0 | T2I + I2I/Editing | 1–3 | Flat $0.03/img (1K/2K) | Speed and cost optimized |
Comparison Table 2: Cross-Model Comparison
| Model | Developer | Pricing | Reference Images | Special Features |
|---|---|---|---|---|
| Qwen Image 2.1 Pro | Alibaba | Flat $0.04/img | Up to 10 | Native RGBA transparent output |
| Seedream 4.0 | ByteDance | Billed per image | Multi-image support | 2K/4K output, multi-image editing |
| GPT Image 2 | OpenAI | Billed per image (> $0.04) | Varies | Strong instruction following, text rendering |
Why Choose Qwen Image 2.1 Pro?
- High Reference Image Limit: Supports up to 10 reference images for complex conditional generation, exceeding the limits of many same-tier models.
- Native Transparency: Natively generates RGBA transparent images without requiring background removal tools.
- Flat Pricing: A uniform $0.04 per image for both 1K and 2K resolutions ensures cost predictability.
- Flexible Resolution: Supports a wide pixel range and aspect ratios, with auto-prediction when size is omitted.
- High-Resolution Output: Capable of direct native 2K resolution output for detailed and high-quality visuals.
Specifications
| Field | Value |
|---|---|
| Model Name | Qwen Image 2.1 Pro |
| Developer | Alibaba (Qwen Team) |
| Model ID | aliyun/qwen-image-2-1-pro |
| Endpoint | https://api.icreat.ai/v1/task/submit/aliyun/qwen-image-2-1-pro |
| Task Result Endpoint | https://api.icreat.ai/v1/task/result |
| Model Type | Image Generation and Editing Model |
| Authentication | Authorization: Bearer ${ICREAT_API_KEY} |
| Input Modalities | Text and Image (Reference URLs) |
| Reference Image Limit | Up to 10 images |
| Input Image Size Limit | Width and height 1–8000 pixels (recommended 384–2048) |
| Output Size Range | Total pixels between 512×512 and 2048×2048 |
| Aspect Ratio Range | 1:8 to 8:1 |
| Output Format | Image (url and download_url) |
| Billing Unit | USD per image (billed per output image; $0.04 flat for 1K and 2K) |
| API Mode | Asynchronous two-step (submit task → poll result) |
Architecture
The API follows an asynchronous two-step architecture. The client submits a generation task via the submit endpoint. The request body consists of two top-level sibling fields: input and parameters. The input field contains a messages array with role and content (including text and image items). The parameters field defines size and watermark options. The request does not use a model field. Upon submission, a task_id is returned. The client then polls the task result endpoint using the same task_id at 2–5 second intervals until the status is SUCCEEDED or FAILED.
Notes
- The submission and result polling steps use the same
task_id. While processing, theresultarray will be empty ([]), and polling should continue. - A FAILED status is terminal; check request parameters, image URLs, or content moderation policies.
- The
inputandparametersfields must be top-level sibling objects and must not be nested within each other. - A maximum of 10 reference images is supported. Input image width and height must be 1–8000 pixels (recommended 384–2048).
- If
sizeis omitted, the model automatically predicts the resolution and aspect ratio. Total pixels must be between 512×512 and 2048×2048, with an aspect ratio of 1:8–8:1. - The
watermarkparameter defaults to false. costUSDis 0.04 (USD) per successful generation and is only returned when the task succeeds.- 1K and 2K tiers share the same billing rate of $0.04/image. Input reference images are not charged additionally.
Frequently Asked Questions
How do I call Qwen Image 2.1 Pro?
The model uses an asynchronous two-step process. First, submit the task to the endpoint https://api.icreat.ai/v1/task/submit/aliyun/qwen-image-2-1-pro using Bearer authentication. Then, poll the https://api.icreat.ai/v1/task/result endpoint with the returned task_id until the task succeeds or fails.
How is billing calculated?
Billing is based on the number of output images generated. Both 1K and 2K resolutions are billed at a flat rate of $0.04 per image, and input reference images are not charged. For example, generating 2 images at 2048*2048 resolution costs $0.08. The costUSD field in the response indicates this amount and is only returned upon successful generation.
How many reference images are supported?
The model supports up to 10 reference images, which can be used for multi-image conditional generation, style transfer, or continuous editing workflows.
What sizes are supported?
The size parameter is customizable. The total pixel count must be between 512×512 and 2048×2048, and the aspect ratio must be between 1:8 and 8:1. If the size parameter is omitted, the model automatically predicts the optimal resolution and aspect ratio based on the prompt.
Does it support transparent backgrounds?
Yes, the model supports native RGBA transparent image generation, allowing direct creation of images with transparent backgrounds.
What is the difference between this and Qwen Image 3.0 Pro?
Qwen Image 2.1 Pro offers a flat rate of $0.04 per image for both 1K and 2K, supports up to 10 reference images, and features native RGBA transparent output. Qwen Image 3.0 Pro charges $0.04 for 1K and $0.075 for 2K, supports 1–3 reference images, and excels in complex text rendering across 12 languages.