Qwen Image 3 API
AI ImageQwen Image 3 is Alibaba's Qwen-Image 3 route on OmniaKey, combining text-to-image generation and reference edits with strong Chinese and English text rendering in one per-call media API.
- Per successful call$0.50
- OperationsText to image
- Image edit
- View tool pricing and billing rules
Your result will appear here once you run the model.
01
Where it fits
Qwen Image 3 is Alibaba's Qwen-Image 3 route on OmniaKey, combining text-to-image generation and reference edits with strong Chinese and English text rendering in one per-call media API.
Choose Qwen Image 3 for efficient generation and edits where accurate on-image text matters, and move to Qwen Image 3 Pro when typography and layout quality lead.
02
What it does
Text to image
qwen-image-3
Turn a text prompt into a new image.
Image edit
qwen-image-3
Change an existing image from a prompt and reference input.
03
In depth
Qwen Image 3 is Alibaba's image model on OmniaKey, listed as qwen-image-3 and mapped to Alibaba's qwen-image-3.0. It turns a text description into a picture and edits an image you supply, both through the asynchronous media task API. Inside the Qwen image line it sits below Qwen Image 3 Pro, which the catalog lists in a higher cost tier per call.
Choose it for work where the wording on the picture has to survive: a concert poster, a menu board, a bilingual shop sign, or a product edit that may not move the product. The ceiling is 2048×2048, so a job that needs 4K belongs with Wan 2.7 Image Pro or Nano Banana 2.1. Prompt rewriting is on by default and rewrites your text; switching it off also switches off the thinking step.
In our same-prompt runs on 25 October 2026 the model spelled every requested English and Chinese word correctly, kept the small poster line in its mixed case, and returned a flat poster rather than a photographed mockup. Its edit left the bottle, cork and label closest to the source photo. On the two photographic prompts it produced a clean, render-like frame with fewer place details than Nano Banana 2.1.
Flat posters with exact wording
Generate a gig poster, menu or flyer where the headline and the small line must match your copy word for word. Qwen Image 3 kept the mixed case of a long small line and returned a full-frame flat file. Send prompt_extend as false so the wording stays literal, and set size to 2048*2048.
Product edits that hold still
Swap a background while the product, its cork and every word on its label stay put. In our test it kept the olive oil name, the sprig drawing and the volume line, and stayed closest to the source photograph. Pass the source in input.images as an array and repeat which details must not change.
Bilingual shop signage
Render a storefront or tea-house plaque that carries two lines of Chinese and one of English. Both Chinese lines came out right in both runs; the risk is the finish, which read as a 3D render. Add a realism phrase and a location detail if the picture has to pass for a photograph.
04
Hands-on notes
In our jazz-poster test, Qwen Image 3 kept the small line in the mixed case and punctuation we had written in the prompt.
Read the testQwen Image 3 was faster than Nano Banana 2.1 on every new image we ran, from 23.1 to 27.4 seconds against 32.0 to 37.1.
Read the testOur Chinese tea-house sign came back correct in both Qwen Image 3 runs, but the finish looked more like a 3D render than a photograph.
Read the testIn our product edit, Qwen Image 3 kept the bottle shape, the cork and every word on the label, and stayed closest to the source photo.
Read the testFor our Lisbon fish-market portrait, Qwen Image 3 returned a usable candid shot, but nothing in the frame read as Lisbon.
Read the test
05
Limits to plan around
- Output stops at 2048×2048, with aspect ratios from 1:8 to 8:1, so a 4K brief cannot be served here. Wan 2.7 Image Pro and Nano Banana 2.1 are the catalog options for larger frames.
- Photographic prompts come back with a clean, render-like finish and thin place cues: our Lisbon street portrait could have come from any indoor fish market. Name the city, the tiles and the light explicitly.
- prompt_extend is on by default and rewrites your description, which undoes careful wording; turn it off, and accept that the thinking step goes off with it.
- An edit takes the source in input.images as an array, so a body written for a model that accepts a single input.image is rejected; send the array even when there is only one file.
06
Compare nearby models
| Model | Best for | Operations | Listed price |
|---|---|---|---|
| Qwen Image 3This page qwen-image-3 | Flat posters, menus and label-safe edits where exact wording matters more than a photographic finish. | Text to image · Image edit | $0.50 per call |
| Qwen Image 3 Pro qwen-image-3-pro | Dense layouts and small type in posters or infographics, where a higher per-call tier buys finer typography. | Text to image · Image edit | $0.60 per call |
| Nano Banana 2.1 gemini-nano-banana-2.1 | Photoreal portraits and street scenes, place detail, and any frame larger than 2048×2048. | Text to image · Image edit | $0.019–$0.039 per call |
| Wan 2.7 Image Pro wan2.7-image-pro | 4K text-to-image work and brand-colour control, with edits capped at 2K. | Text to image · Image edit | $0.99 per call |
07
Sources and verification
Guide updated 2026-10-27.
08
Call it
Media models run asynchronously. Create a task, poll it until it finishes, then download the signed asset URL.
# 1. Create the job. 202 Accepted; the body carries the task id,
# the pre-charged price, and a poll_url. Reuse one Idempotency-Key
# across retries so a timeout never starts a second paid job.
curl --fail-with-body https://api.omniakey.com/v1/media/tasks \
-H "Authorization: Bearer $OMNIAKEY_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: qwen-image-3-001" \
-d '{
"model": "qwen-image-3",
"type": "image",
"operation": "generate",
"input": {"prompt": "a paper boat on a quiet lake at sunset"},
"parameters": {"size": "1024x1024", "quality": "medium"}
}'
# 2. Poll GET /v1/media/tasks/{id} until status is
# succeeded | failed | canceled. Progress carries percent and stage.
curl --fail-with-body https://api.omniakey.com/v1/media/tasks/TASK_ID \
-H "Authorization: Bearer $OMNIAKEY_API_KEY"
# 3. Download assets[0].url before expires_at, or poll again for a
# fresh signed URL. Failed and canceled jobs are refunded.
curl -sS -o out.png "$(jq -r '.assets[0].url' result.json)"09
Pricing
Image jobs are billed per successful call, not per token. Failed or canceled jobs are refunded automatically.
$0.50per call
More Alibaba image models
Related articles
Frequently asked questions
How do I call Qwen Image 3 on OmniaKey?
POST to /v1/media/tasks with the model id, type "image", and operation "generate" or "edit". Reuse one Idempotency-Key across retries so a timeout never starts a second paid job.
What does Qwen Image 3 cost?
Qwen Image 3 bills $0.50 per call. Creation pre-charges that amount from your prepaid balance, and failed or canceled jobs are refunded automatically.
Why does the API return a task id instead of the image?
Provider latency runs from seconds to about a minute, so OmniaKey normalizes every image route into one asynchronous lifecycle. Poll GET /v1/media/tasks/{id} until the status is succeeded, then read the signed asset URL.
Can Qwen Image 3 make 4K images?
No. Its largest output is 2048×2048 and the aspect ratio runs from 1:8 to 8:1, so 2K is the ceiling. Wan 2.7 Image Pro accepts 4K text-to-image requests and caps its edits at 2K; Nano Banana 2.1 also offers a 4K size. For anything bigger, generate at 2048 and upscale afterwards.
Is Qwen Image 3 open source?
No. The Qwen team's Hugging Face page lists the Qwen-Image-2.1 family and no weights for 3.0. The newest open Qwen image model is Qwen-Image-2.1, a 7B model under the Qwen Research License. Running those weights on your own machine is a different job from calling hosted qwen-image-3.
How is Qwen Image 3 billed?
OmniaKey bills it per call, and there is no token charge on top. The rate is the one in the price table on its model page, and a failed task is refunded. A generate and an edit cost the same per call.
How do I send a request for Qwen Image 3?
Create a task with POST https://api.omniakey.com/v1/media/tasks, set model to qwen-image-3, type to image, operation to generate or edit, and put the description in input.prompt. Add an Idempotency-Key header, then poll GET /v1/media/tasks/{id} until status reads succeeded and fetch assets[0].url before that signed link expires.
Why did my wording come back changed?
prompt_extend is on by default and rewrites the prompt before the model draws, so a carefully typed line can come back reworded or recased. Set prompt_extend to false to use your exact text, which also disables the thinking step; our poster and tea-house runs were sent that way.
Does Qwen Image 3 add a watermark?
Only if you ask for one. The watermark parameter defaults to false and our images carried no label. If you want a visible mark, set watermark to true on the request; the setting applies to both generate and edit operations.