Access Qwen Image 3.0 and Qwen Image 3.0 Pro—also searched as Qwen3 Image or Tongyi Wanxiang Image 3.0—through EvoLink's unified image API. Test text-rich layout generation, 1-3 reference-image editing, and 1-6 image batches before integrating.
Test Qwen Image 3.0 output quality, estimate credit usage, and generate a sample before you formally integrate.
Create image task
Fixed to qwen-image-3.0-proAdd up to 3 reference images for image-to-image. Leave empty for text-to-image.
The model auto-selects the resolution from your prompt.
Results are kept for 24 hours — download anything you want to keep.
EvoLink bills this route per generated image at the current account price. Capacity may vary while upstream access remains limited.
First Test Cost
Standard · 1K · 1 image, no input imagesOne generated image
About 302 images with $10 in credits.
In Playground, select Standard, 1K and 1 image with no references for this price. Output is billed per generated image at the 1K or 2K tier. Reference images are billed separately, so image-to-image costs more than text-to-image.
Testing Budget Guide
Pick an amount based on how many images you expect.Good for a first validation.
Good for prompt iteration.
Good before production integration.
Estimated cost for generated images. Final cost is based on the number actually generated.
| Model tier | Item | Rule | Rate | Billing |
|---|---|---|---|---|
| Qwen Image 3.0 Pro | Output image - 1K | Generated image with a total area of 2,250,000 pixels or less. | $0.045/image2.9977 cr/image | Per image (× n) |
| Qwen Image 3.0 Pro | Output image - 2K | Generated image larger than 2,250,000 pixels in total area. | $0.083/image5.6207 cr/image | Per image (× n) |
| Qwen Image 3.0 Pro | Reference image | Each reference image sent for image-to-image editing (up to 3). | $0.0034/image0.2248 cr/image | Per reference image |
Total = reference images x input rate + generated images x output tier rate. Text-to-image has no reference-image charge. With size=auto the tier is unknown at submit time, so credits are held at the 2K rate and the difference is refunded on settlement.
Use it when a single brief must coordinate exact copy, multiple layout zones, multilingual typography, reference roles, and a final delivery format. The practical question is not whether one demo looks good, but whether the model lowers revision work and cost per accepted image for your workflow. EvoLink keeps generation, editing, billing, evaluation, and fallback routes in one account so teams can answer that question with production-like tests.
qwen-image-3.0-proHighest fidelity tier with a separate 2K output rate — the default identity of this page for text-heavy production visuals.
qwen-image-3.0Standard tier with a lower per-image rate, and 1K and 2K priced the same — same request contract as Pro.
modelstringSelect Standard or Pro through the model ID.
image_urlsstring[]Omit for text-to-image, or send 1–3 public image URLs for editing.
nintegerGenerate 1–6 images; the default is 1.
qualityenumChoose 1K or 2K when using a ratio size; custom pixel dimensions determine the billing tier.
Use the EvoLink API for product integration and batch jobs, or call it from Codex, Claude, or Gemini for fast creative and development workflows. Both paths share the same EvoLink API key, balance, model routes, and task history.
Best for: product backends, batch jobs, automated workflows
Call EvoLink’s unified image API from your server and control the model ID, parameters, task queue, callbacks, and result storage.
Best for: creative and development tasks in Codex, Claude, and Gemini
Give the Agent your output goal, assets, and acceptance criteria. It can choose the route, assemble the request, track the task, and return the result without requiring you to hand-code every step.
This example shows the shortest runnable flow: create a task, save the returned task ID, then poll or wait for an HTTPS callback. Open the API tab for the complete parameter and response reference.
curl -X POST https://api.evolink.ai/v1/images/generations \
-H "Authorization: Bearer $EVOLINK_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen-image-3.0",
"prompt": "A product shot of a ceramic mug on a wooden table, soft daylight",
"size": "16:9",
"quality": "1K",
"n": 1,
"callback_url": "https://your-domain.com/webhooks/image-task-completed"
}'
# Save the returned task ID, then query:
curl https://api.evolink.ai/v1/tasks/{task_id} \
-H "Authorization: Bearer $EVOLINK_API_KEY"Check every value against the parameter cards above: ranges, allowed options, asset counts, and file types must match the selected route.
Check the Authorization bearer token and confirm the available balance and task billing in the console.
Review likeness rights, brand assets, readable text, sensitive content, and input-asset compliance.
A Qwen Image 3.0 task that reaches a final failed state is not charged. Keep the task ID, inspect the final state, and identify whether the cause is validation, review, or execution before retrying.
callback_url must be a public HTTPS URL, not localhost or a private network address. Always keep the task ID as a polling fallback.
| Decision factor | Qwen Image 3.0 | Nano Banana | GPT Image |
|---|---|---|---|
| Primary strength | Text-rich complex layouts | Flexible creative generation | Generation and editing workflows |
| Prompt depth | Up to 4.5K tokens | Model dependent | Model dependent |
| Text and layout | 10px text; 12 languages | Strong general rendering | Strong general rendering |
| Reference input | 1-3 images | Model dependent | Supported |
| Route stage | Live | Varies by model | Varies by model |
Send only a prompt for text-to-image, or include 1-3 image references for image-to-image and guided editing workflows. The 4.5K-token input window supports detailed layout, typography, style, color, object, and knowledge constraints in one brief. Name the artifact, audience, canvas, zones, exact copy, visual rules, invariants, and acceptance criteria. Add length only when it defines a real relationship or review rule. State whether each image controls the subject, identity, pose, composition, palette, lighting, or style, and list what must remain unchanged. Ambiguous multi-reference prompts often blend the wrong attributes.
Keep the returned task ID. Poll /v1/tasks/{task_id} or use callback_url for the final state; download results within 24 hours. The EvoLink route is live, but upstream access remains limited. Start with a controlled workload and measure queue behavior, latency, output acceptance, retries, and capacity before moving critical traffic.
Output fees follow the selected tier and the number of images generated. Reference images are charged separately. Failed tasks are refunded. Treat small-text readability, spelling, layout, formulas, chart relationships, and factual accuracy as separate review gates. A polished knowledge graphic can still contain a wrong label or invented fact.
Use one EvoLink API key for Qwen and other image, video, audio, and language model routes.
Test Qwen beside other image routes without creating a separate vendor account, billing setup, and integration stack.
Use one balance and one console to compare spend across test and production image workloads.
Track submitted, processing, completed, and failed image tasks through the same EvoLink workflow.
Keep alternative image routes available for workload-based selection, cost comparison, and failover.
Compare Qwen image routes under one EvoLink account and API key, then choose by editing depth, reference workflow, and cost.
Qwen Image 3.0
CurrentUse it when a single brief must coordinate exact copy, multiple layout zones, multilingual typography, reference roles, and a final delivery format. The practical question is not whether one demo looks good, but whether the model lowers revision work and cost per accepted image for your workflow. EvoLink keeps generation, editing, billing, evaluation, and fallback routes in one account so teams can answer that question with production-like tests.
Qwen Image Edit Plus
Use Qwen Image Edit Plus when the job centers on multi-reference editing rather than long, text-heavy generation briefs.
View API
Qwen Image Edit
Use the established Qwen Image Edit route for focused single-image transformations and localized creative changes.
View API
Seedream 5.0 Pro
Use Seedream 5.0 Pro for high-fidelity generation and multi-reference creative workflows.
View API
Nano Banana Pro
Use Nano Banana Pro when high-fidelity general image generation is the priority.
View API
Compare capability gains, limited-access and capacity risks, evaluation metrics, and the workloads that should migrate first.
Read comparisonCreate the first task, poll asynchronous status, store results, and add retry and fallback controls.
Read API guideBuild structured prompts, assign reference roles, review exact text and facts, and troubleshoot common layout failures.
Read prompt guideCompare text-heavy layouts, general creative work, editing fit, access maturity, and a workload-based evaluation plan.
Read model comparisonYes. The route is live on EvoLink. Upstream access remains limited, so verify capacity and behavior with your own workload before moving critical production traffic.
Use qwen-image-3.0-pro with the EvoLink image generation endpoint shown in the API tab.
The 3.0 release emphasizes richer content, more authentic detail, and deeper knowledge, including up to 4.5K-token prompts, 10px text, and native rendering in 12 languages.
Yes. Add 1-3 reference images to guide an image-to-image edit, or omit references for text-to-image generation.
The n parameter supports 1-6 generated images per request. Cost follows the number of successful outputs.
Cost depends on the selected model tier, output size, number of generated images, and any reference images. Check the live Pricing tab for the current account rate and billing breakdown.
Four ways to set size: auto (the model picks a resolution), a quality tier (1K/2K), an aspect ratio such as 1:1 or 16:9, or custom WIDTHxHEIGHT pixels. Custom pixels use total pixels 262,144–4,194,304 with aspect ratio 1:16–16:1, and size works with any n (1–6).
prompt_extend lets the model expand a short prompt before generation. Use it for exploration, but keep it controlled when exact wording and layout instructions matter.
Both tiers use the same EvoLink image workflow and parameter contract. Standard is the lower-cost option; Pro is the higher-fidelity tier and uses a separate 2K output rate. Test both with the same acceptance set before choosing a default.
Treat 10px text as a vendor-stated capability, not a guarantee that every character will be correct. Review spelling, punctuation, formulas, contrast, and readability at the final delivery size after resizing or compression.
Qwen states that the model natively renders 12 languages, but a headline claim does not replace locale QA. Build a test set for the exact scripts, punctuation, fonts, brand terms, and mixed-language layouts your product will publish.
No. It helps the model express knowledge-rich formats, but generated facts, equations, chart relationships, dates, labels, and medical or scientific content still require source-based human review.
Start with the artifact and audience, then define canvas and zones, exact quoted copy, visual style, reference roles, invariants, and an acceptance checklist. Do not turn the full 4.5K-token allowance into one unstructured paragraph.
Assign one explicit role to each image, such as subject, pose, composition, or style. State which source attributes must be preserved and which may change; then review identity and untouched regions separately.
No official Qwen Image 3.0 weights, model card, or self-hosting license were verified in the current official release material. Do not plan a local deployment until Qwen publishes the artifacts and terms.
Record model tier, prompt, references, parameters, output size, queue and generation latency, retries, credits, acceptance reason, and fallback use. Compare cost per accepted image, not only the listed price per generated image.
The EvoLink route is live, but production readiness is workload-specific. Start with a canary, then expand only after measuring latency, acceptance rate, queue behavior, retries, and capacity.
Keep a fallback when your workflow prioritizes mature capacity, a different editing style, or provider-specific features over Qwen Image 3.0’s text-rich layout strengths.