GPT Image 2.5 is live — OpenAI's newest image model, targeted edits that leave the rest of the frame alone
All Posts
Qwen Image 3.0 API Pricing: Pro vs Standard Cost per Image
2026/10/07

Qwen Image 3.0 API Pricing: Pro vs Standard Cost per Image

Qwen Image 3.0 API pricing explained: Alibaba's per-image rates for Standard and Pro at 1K and 2K, how reference images are billed, and when Pro is worth it.

Qwen Image 3.0 API pricing is per image, not per token. On Alibaba Cloud Model Studio's Singapore region, qwen-image-3.0 costs $0.03 per image at both 1K and 2K, while qwen-image-3.0-pro costs $0.04 at 1K and $0.075 at 2K, with $0.003 for every input image you send for editing[1]. Those are Alibaba's list prices as published on October 7, 2026, and other regions use a lower rate card.

The structure matters more than the exact cents, because it decides what a workload costs: on Standard, resolution is free; on Pro, 2K roughly doubles the bill; and reference images are charged per request on top. This guide covers Alibaba's rate card, the billing dimensions on reAPI, when the Pro tier earns its premium, and the shortest path from an API key to a first image.

TL;DR

  • Alibaba Singapore list prices: Standard $0.03/image at 1K and 2K; Pro $0.04 at 1K and $0.075 at 2K; input images $0.003 each[1].
  • Other Alibaba regions (Beijing, Hong Kong, Frankfurt, Virginia, Tokyo): Standard $0.024754, Pro $0.034380 at 1K and $0.068761 at 2K, input images $0.00275[1].
  • Failed images are not billed on Alibaba[1] or on reAPI[2].
  • On reAPI, Standard is one rate for 1K and 2K, Pro 2K costs twice Pro 1K, and each reference image adds a small charge once per request. Current rates are on the Qwen Image 3.0 model page[3].
  • Pro is worth it for dense text and layout. On general prompts, the two tiers sit 13 Elo points apart on Artificial Analysis's arena[4].

What Alibaba charges for Qwen Image 3.0

Alibaba bills Qwen image models "based on the number of input images and the number of successfully generated images", using this formula[1]:

Cost = Input image unit price × Number of input images + Image unit price × Number of images generated

The Singapore (International) rate card, as listed on October 7, 2026[1]:

Model idOutput 1KOutput 2KPer input imageFree quota
qwen-image-3.0$0.03$0.03$0.00310 images
qwen-image-3.0-pro$0.04$0.075$0.00310 images

Beijing, Hong Kong, Frankfurt, Virginia and Tokyo share a different card[1]:

Model idOutput 1KOutput 2KPer input image
qwen-image-3.0$0.024754$0.024754$0.00275
qwen-image-3.0-pro$0.034380$0.068761$0.00275

Three details are easy to miss. The free quota exists only in Singapore and is valid for 90 days from Model Studio activation, model release or application approval, whichever is later[1]. The Pro 2K premium is not the same everywhere: $0.075 / $0.04 = 1.875× in Singapore, but $0.068761 / $0.034380 = 2.0× in the other regions. And failed requests "incur no cost and do not consume your free quota"[1].

A worked example on the Singapore card: one Pro edit at 2K with two reference images costs 2 × $0.003 + 1 × $0.075 = $0.081. The same request on Standard costs 2 × $0.003 + $0.03 = $0.036.

How Qwen Image 3.0 API pricing works on reAPI

On reAPI the model ids are the same, qwen-image-3.0 and qwen-image-3.0-pro, and the bill has the same shape. What moves the price[2][3]:

DimensionEffect on the bill
TierPro costs more per image than Standard
ResolutionNo effect on Standard; on Pro, 2K costs twice 1K
n (1–6)Every delivered image is billed
Reference images (image_urls, 1–3)Each distinct URL adds a small charge, once per request, not once per output image
Raw pixel sizeAn area above 2.25 million pixels bills at the 2K rate, even without resolution: "2K"

What does not move the price: prompt_extend, negative_prompt, the aspect ratio, and the content filter setting. Polling the task is free, and a failed generation refunds its reserved credits automatically[2].

reAPI bills in credits, where 1 credit = $0.001, and the total is computed as[2]:

credits = ceil((per_image_usd × n + per_reference_usd × references) × 1000)

Because the reference charge is per request, a batch of n: 4 with two references pays for four images and two references, not eight references. Duplicate URLs count once. The per-image and per-reference rates for both tiers are listed on the model page, which reads them from the live price table, and the playground shows an estimate before you submit. There is no subscription; credits never expire.

Standard vs Pro: when Pro is worth paying for

Alibaba positions the two tiers clearly. Its model guide recommends qwen-image-3.0-pro for "complex layouts such as newspapers, storyboards, menus, and exam papers" and small text down to 10 pixels, and describes qwen-image-3.0 as "a faster version of qwen-image-3.0-pro". Both accept the same inputs, return up to six images and reach 2048×2048[5].

Independent preference data puts the quality gap in perspective. On Artificial Analysis's text-to-image arena on October 7, 2026, Qwen-Image-3.0-Pro scored an Elo of 1088 ±9 and Qwen-Image-3.0 scored 1075 ±9, based on about 6,600 and 6,300 votes[4]. The confidence intervals nearly touch. Arena prompts are mostly general images, so that small gap says little about dense typography, where Pro is supposed to differ.

Our reading of that evidence, as an editorial judgment rather than a measured result:

  • Use Standard by default for illustrations, photos, product shots and social images. At 2K it costs the same as at 1K, so it is the cheapest way to get 2K pixels from Qwen Image 3.0.
  • Use Pro when the image has to carry a lot of readable text: menus, newspaper pages, exam papers, infographic grids, UI mockups. Run a few of your real prompts on both tiers first; if Standard already renders the text cleanly, Pro is not buying you anything.
  • On Pro, draft at 1K. Since 2K costs twice 1K, iterate on composition at 1K and re-run only the winner at 2K.
  • Keep reference images to the ones you need. Each extra reference is billed, though the charge is small next to the image itself.

For how Qwen Image 3.0 compares with other models on price and output, see Qwen Image 3 vs GPT Image 2.

Getting an API key and running a first request

  1. Sign up at reAPI and open API keys in the dashboard to create a key. Eligible new accounts receive signup credits for testing.
  2. Send a request. This one uses the Pro tier at 1K for a layout-heavy prompt:
curl https://reapi.ai/api/v1/images/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen-image-3.0-pro",
    "prompt": "A cafe menu board, three columns - appetizers, mains, desserts - five dishes each with prices, serif headings, cream paper background",
    "size": "3:4",
    "resolution": "1K"
  }'
  1. The response contains a task id. Poll GET /api/v1/tasks/{id} until status is completed, then read output.image_urls.

To edit instead of generate, add "image_urls" with one to three public https image URLs; base64 and data: URIs are rejected. The API docs list every parameter, including negative_prompt, prompt_extend and explicit pixel sizes, with the same request in Python, Node.js and Go[2].

FAQ

Qwen image 3.0 cost

On Alibaba Cloud Singapore, Qwen Image 3.0 costs $0.03 per image at 1K or 2K, and Qwen Image 3.0 Pro costs $0.04 at 1K and $0.075 at 2K, plus $0.003 per input image[1]. reAPI's current rates for both tiers are on the model page.

How many images can I generate with Qwen?

Up to six images per request with Qwen Image 3.0, on Alibaba Cloud[5] and on reAPI (n from 1 to 6)[2]. Each delivered image is billed. Alibaba's Singapore free quota covers 10 images per 3.0 model[1].

Qwen image 3 pro free

Not as a free plan. Alibaba's only free allowance is 10 images of qwen-image-3.0-pro in the Singapore region, valid for 90 days[1]. Qwen Image 3.0 has no downloadable weights to self-host either; see Is Qwen Image 3 open source?.

Qwen image 3.0 free API key

API keys are free to create on Alibaba Cloud and on reAPI; the images are what you pay for. On reAPI, eligible new accounts receive signup credits, which is enough to try a few generations before topping up.

Qwen image 3.0 API python

Send the same JSON body with requests.post to https://reapi.ai/api/v1/images/generations with an Authorization: Bearer header, then poll the task endpoint. A complete Python example is in the API docs[2].

Is the Qwen image any good?

Third-party preference voting rates it well. On Artificial Analysis's text-to-image arena, Qwen-Image-3.0-Pro (1088) and Qwen-Image-3.0 (1075) ranked 14th and 16th in the current-models view on October 7, 2026[4]. Alibaba's own announcement of 3.0 includes no benchmark scores; it focuses on long layout prompts and small text[6].

How to use qwen image 3

Get an API key, POST a prompt with "model": "qwen-image-3.0" or "qwen-image-3.0-pro" to the images endpoint, and poll the returned task until it completes. Add image_urls to edit an existing image. You can also try prompts in the playground on the model page before writing code.

Budgeting a Qwen Image 3.0 workload

Start from the dimension that dominates your volume. For high-volume general images, Standard is the cheaper tier and its flat price makes 2K free. For text-heavy layouts where Pro's typography matters, budget Pro at 1K for drafts and pay the 2K premium only on finals. Count reference images per request, not per output.

Qwen Image 3.0 API pricing on reAPI follows that same per-image, per-reference structure, with failed generations refunded and no subscription. Check the live rates on the Qwen Image 3.0 model page and the full parameter list in the API docs.

References

  1. Alibaba Cloud. Model Studio: Model inference pricing (Qwen Image Generation and Editing). Retrieved October 2026 from alibabacloud.com/help/en/model-studio/model-pricing
  2. reAPI. Qwen Image 3.0 API docs. reapi.ai/docs/qwen-image-3
  3. reAPI. Qwen Image 3.0 model page. reapi.ai/models/qwen-image-3
  4. Artificial Analysis. Text to Image Leaderboard. Retrieved October 2026 from artificialanalysis.ai/image/leaderboard/text-to-image
  5. Alibaba Cloud. Model Studio: Image generation and editing. Retrieved October 2026 from alibabacloud.com/help/en/model-studio/image-model
  6. Qwen Team. Qwen-Image-3.0: Rich Content, Authentic Details, Deep Knowledge. Retrieved October 2026 from qwen.ai/blog?id=qwen-image-3.0
All Posts

More Posts