Skip to main content
AI image model
Alibaba Cloud
Qwen Image

Qwen Image 3.0

Qwen Image 3.0 supports both text-only generation and image-to-image work with up to 3 references. It offers 1K/2K, 8 canvas sizes, PNG/JPEG output, prompt expansion, negative prompts, and seed control. Kyeo preauthorizes 5–7 credits per request.

Workflows
Text to image, image to image
Resolution
1K/2K
References
Up to 3 images
Credits
5–7 per request
Uses the appropriate text-to-image or image-to-image entry point for 1–3 references
1K/2K share the same base price, with text-only generation preauthorized at 5 credits
8 canvas sizes with PNG/JPEG output
Supports prompt expansion, negative prompts, a sensitive-content toggle, and seed
30-second overview
Qwen Image
5–7 credits
What it does best
A general image-generation model for controlled canvases, reproducible random seeds, and a small set of multiple references.
Best for
Marketing drafts, character or product references, horizontal and vertical content, visual exploration with negative constraints, and delivery at 1K/2K.
Popular searches
how to use Qwen Image 3.0how much does Qwen Image 3.0 costQwen Image 3.0 reference image limit

At a glance

What is this model like?

Without uploads, the model uses its text-to-image entry point; uploading 1–3 images switches to image to image. References may be JPEG, PNG, WebP, BMP, GIF, or TIFF files up to 10 MB each. The client submits only fields allowed by the current contract, and extra parameters are rejected before charging.

What it does best
A general image-generation model for controlled canvases, reproducible random seeds, and a small set of multiple references.
Best for
Marketing drafts, character or product references, horizontal and vertical content, visual exploration with negative constraints, and delivery at 1K/2K.
Why use it on Kyeo AI
Kyeo rounds preauthorization up based on the number of input images and locks the text-to-image or image-to-image route, image formats, and parameter allowlist before charging.

Key facts

Quickly assess whether this model fits your use case.

Model category
AI image model
Vendor
Alibaba Cloud
Model family
Qwen Image
Prompt limit
800 characters
Reference formats
JPEG/PNG/WebP/BMP/GIF/TIFF
Output formats
PNG/JPEG

Also known as

The same model may appear under different names across documentation and community discussions; this list keeps them easy to search and compare.

Qwen Image 3
Qwen3 Image
Alibaba Qwen Image 3.0
Qwen image generation 3.0

Common questions

These practical questions focus on the task, input conditions, and result requirements you should confirm before choosing.

how to use Qwen Image 3.0
how much does Qwen Image 3.0 cost
Qwen Image 3.0 reference image limit
Qwen Image 3.0 supported sizes
Qwen Image 3.0 vs Pro
how to set seed and negative prompts in Qwen Image 3.0

Selection guide

Use these decision points when choosing a model.

1
Choose the standard model to control cost

If you want the same base price at 1K/2K, the standard model is easier to budget than Pro. Preauthorization still rises according to the reference-image formula.

2
Compare Pro for the higher-priced tier

Pro has a higher 2K base price, so treat it as a separate pricing tier rather than a free upgrade from the standard model.

3
Switch models for a larger reference set

The current image-to-image limit is 3 images. For 4 or more, compare a multi-image entry point such as Nano Banana 2 Lite.

Popular comparisons

Compare common alternatives on the same task to clarify differences in inputs, controls, and cost.

How do Qwen Image 3.0 and Qwen Image 3.0 Pro differ in price and resolution?
What control changes does Qwen Image 3.0 introduce over Qwen2 Image?
How are multiple references priced in Qwen Image 3.0?

Model comparison

Compare the current model with alternatives at a glance.

Kyeo credits
Qwen Image 3.0
5–7 credits
Qwen Image 3.0 Pro
7–14 credits
Nano Banana 2 Lite
Fixed 3 credits
Resolution
Qwen Image 3.0
1K/2K at the same base price
Qwen Image 3.0 Pro
1K/2K with tiered pricing
Qwen2 Image
See the current contract on its model page
References
Qwen Image 3.0
1–3 images, 10 MB each
Qwen Image 3.0 Pro
1–3 images, 10 MB each
Nano Banana 2 Lite
Up to 10 images, 30 MB each

Practical usage insights

Practical guidance based on public sources, current on-site limits, and representative tasks.

Price multiple references by actual count

Each input adds 0.5 credits before Kyeo rounds up to an integer. Requests with 1 or 2 references therefore preauthorize 6 credits, while 3 references use 7 credits.

More controls do not make outcomes certain

The seed parameter can help reproduce test conditions, but references, prompt expansion, and asynchronous service state still influence the result. Review final assets manually.

Capabilities

Two-mode image generation

Selects text to image or image to image based on the presence of references while keeping each mode's inputs separate.

Multi-dimensional output controls

Combine resolution, canvas size, output format, prompt expansion, a negative prompt, and a random seed.

Multiple reference formats

Accepts 6 common image formats, with server-side checks of the true file type rather than the extension alone.

Use cases

Product and character references

Use 1–3 images to control subject, material, or composition separately, and explain the purpose of each reference in the prompt.

Drafts for multiple placements

Switch among 1:1, 16:9, 9:16, and 21:9 to prepare directions for covers, feeds, and wide layouts.

Reproducible experiments

Fix the seed, then adjust only the negative prompt or prompt-expansion toggle to compare the effect of one variable.

Prompt tips

Put hard constraints first within 800 characters

State the subject, composition, camera, and required preserved elements before adding style, so critical requirements do not get buried at the end.

Keep negative prompts specific

Name unwanted objects, defects, or layout issues instead of stacking generic low-quality terms.

Assign each reference a role

Identify which image controls the person, color, or spatial relationship to reduce semantic conflict across inputs.

Why choose it

1K/2K share one base price, keeping standard-tier costs clear.
Supports 3 reference images in 6 input formats.
Canvas, format, negative-prompt, and seed controls are comprehensive.
Input count is included in the price calculation before charging instead of preauthorizing only the minimum.

What to know first

Prompts are limited to 800 characters, so complex specifications must be concise.
Image-to-image tasks require 1–3 images, each no larger than 10 MB.
References add cost: 3 images reach a 7-credit preauthorization.
Asynchronous output can still be affected by queues, content review, and public source service state.

Adjustable parameters

Quickly assess whether this model fits your use case.

Resolution
resolution
Optional
Parameter type: Select
Default: 1K
1K
2K
Canvas size
image_size
Optional
Parameter type: Select
Default: 16:9
1:1
3:2
2:3
4:3
3:4
16:9
9:16
21:9
8 options total
Output format
output_format
Optional
Parameter type: Select
Default: png
png
jpeg
Prompt expansion
prompt_extend
Optional
Parameter type: Toggle
Default: on
Sensitive-content toggle
nsfw
Optional
Parameter type: Toggle
Default: off
Negative prompt
negative_prompt
Optional
Parameter type: Text
Up to 5,000 characters
Random seed
seed
Optional
Parameter type: Integer
Default: 1
0–2,147,483,647

Credit usage

5–7 credits

The current public source price is 4.8 credits per output image plus 0.5 credits for each input reference. Kyeo rounds preauthorization up: text-only generation uses 5 credits, while requests with 1/2/3 references use 6/6/7 credits. The base price is the same for 1K and 2K, and final settlement reads actual consumption after completion.

Budget tip

Before batch generation, run an A/B test with the same assets across the current model and alternatives to validate quality and avoid wasting credits.

FAQ

Related models

Compare these similar candidates before deciding.

Qwen Image 3.0 Pro

Qwen Image 3.0 Pro offers text-to-image generation and image to image with up to 3 references. It includes 1K/2K, 8 canvas sizes, PNG/JPEG output, prompt expansion, negative prompts, and seed control. Preauthorization ranges from 7–14 credits based on resolution and input count.

AI image model
7–14 credits

Qwen2 Image

Qwen2 Image — Qwen2 Image / Text creation and one-image editing — Start from a text prompt or upload exactly one image to edit. Multiple reference images are not supported. Ratio, format, and seed controls — Text creation supports 1:1, 3:4, 4:3, 9:16, and 16:9. Editing adds 2:3, 3:2, and 21:9; the default remains 16:9. Prompts are capped at 800 characters; editing accepts one JPG, PNG, or WebP image up to 10 MB; each result costs a fixed 6 credits.

AI image model
Fixed 6 credits per image

Nano Banana 2 Lite

Nano Banana 2 Lite is a streamlined image entry point: generate from text alone or upload up to 10 reference images, choose only the aspect ratio, and use a fixed 3 credits per request. There is no resolution button, which prevents a pricing-table label from being presented as an interface capability.

AI image model
Fixed 3 credits

Sources

Content is based on model documentation, feature references, and the settings available on Kyeo AI.

Source note

Explore Qwen Image 3.0 text-to-image and image-to-image generation, 1K/2K, 8 canvas sizes, up to 3 references, prompt expansion, negative prompts, and 5–7-credit pricing. The current public source price is 4.8 credits per output image plus 0.5 credits for each input reference. Kyeo rounds preauthorization up: text-only generation uses 5 credits, while requests with 1/2/3 references use 6/6/7 credits. The base price is the same for 1K and 2K, and final settlement reads actual consumption after completion.

Last updated: 2026-08-11
Qwen Image 3 current interface notes
Platform

Used to check the input modes, visible controls, media limits, and availability of the version currently connected to this site. The page and workbench show the actual options.