Skip to content

GPT Image API reseller: what to check before you pick one

A GPT Image API reseller, also called a relay or proxy, sells access to image models through its own endpoint and its own billing. Cheapest Imagen is one: you send OpenAI-style requests to our base URL, pay $0.008 or more per image with a fixed price, and nothing is charged when a request fails.

Checked on 2026-11-19

What a reseller or relay API is

You normally reach GPT Image through the maker’s own API: you open an account with the maker, pass its checks, and pay its price. A reseller sits in between. It gives you its own endpoint, its own API key and its own balance, and it passes your request on to the model.

Most resellers copy the maker’s request format, so existing code and SDKs work after you change two things: the base URL and the key. That is why people also call it a relay, a proxy or a third-party API. The words mean the same thing here.

In our case, the part that makes it cheaper is stated on the about page: we buy capacity in bulk and sell it per image with a small margin. We are an independent service and not affiliated with OpenAI or Google.

How it differs from calling the maker directly

The maker’s own APIA reseller such as Cheapest Imagen
AccountAn account with the maker; OpenAI’s GPT Image models need a verified organizationAn account with the reseller, by email
PriceBy token, so it changes with quality and sizeUsually a fixed price per image
ModelsThe maker’s models onlyOften several makers behind one key
Request formatThe maker’s ownOften the same; you change the base URL and the key
Who you pay and who supports youThe makerThe reseller

Compare the price with the maker’s own price table: for OpenAI it is the cost calculator in its image generation guide.

The trade is that you rely on a second company. That is the reason to read the six points below before you put one into production.

What to look for when you pick one

  1. What the price covers. A price per image is easy to compare; a price per token is not. Check which quality and resolution the number is for, and whether it changes when you ask for 2K or 4K. Then compare it with the maker’s price for the same image.
  2. What happens when a request fails. Ask whether a failed request, a timeout or a refused prompt is charged. A low price that bills you for failures can cost more than a higher one that does not.
  3. Whether the API is really compatible. Test with the official SDK, changing only the base URL and the key. Then check the edges: which parameters return an error, whether edits with reference images work, and which output formats you get.
  4. Rate limits. Look for the number of requests per minute and the number running at once, and whether they are written down.
  5. Where your data goes. Find out what is stored, for how long, and whether the prompt is changed on the way. Your prompts and images pass through the reseller, so this matters more than with a direct account.
  6. Whether you can get money back. Find the refund policy before you top up, and check whether unused credits expire.

How Cheapest Imagen answers those points

PointWhat Cheapest Imagen does
PriceA fixed price per image for each model, whatever the quality or resolution. GPT Image 2, 2.5 and 2.6 (1K) cost $0.008 (8 credits); the flare versions of 2.6 cost $0.02; the sunburst versions $0.025. These are for the $100 pack; in the $10 pack GPT Image 2 is $0.014.
Against the makerAt high quality, 1K, that is 96% below the official price for GPT Image 2 ($0.211) and 84% below it for GPT Image 2.6 ($0.053). OpenAI’s low quality setting costs less than ours; see the alternative page.
FailuresA failed request, a timeout or an image the model refuses costs nothing. The exception is the stable channels, where a refused image is still charged.
CompatibilityThe OpenAI Images API format: POST /v1/images/generations and /v1/images/edits. Masks, streaming and transparent backgrounds are not supported. The details are in Switch from the OpenAI API.
Limits300 requests a minute and 50 running at once per account.
Your dataThe prompt is passed to the model unchanged; we do not rewrite it. With response_format set to "url" the image is stored and the link works for 72 hours; by default the image is returned in the response and not stored.
RefundsUnused paid credits are refunded on request. Paid credits never expire. See the refund policy.

Try it with the SDK you already use

New accounts get 5 free images (1K, GPT Image 2, 2.5 and 2.6) to use within 7 days, on the site or through the API, so you can test compatibility before you pay anything.

from openai import OpenAI

client = OpenAI(base_url="https://api.cheapestimagen.com/v1", api_key="YOUR_KEY")
image = client.images.generate(
    model="gpt-image-2.6",
    prompt="a lighthouse on a cliff at dusk, oil painting",
    size="1024x1024",
    response_format="url",
)
print(image.data[0].url)
curl https://api.cheapestimagen.com/v1/images/generations \
  -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "gpt-image-2.6", "prompt": "a lighthouse on a cliff at dusk", "size": "1024x1024", "response_format": "url"}'

The base URL is https://api.cheapestimagen.com/v1. Create the key on the API keys page. Every model, its resolutions and its price are on the models page, and the packs are on the pricing page. For a comparison with other sellers, see the cheapest image generation API.

Questions

What is a GPT Image API reseller?

A company that gives you its own endpoint, key and balance for GPT Image, and passes your requests on to the model. Most copy the OpenAI request format, so your code changes only the base URL and the key.

Is a relay API the same as a proxy?

In practice, yes. Relay, proxy, reseller and third-party API all describe a service between your code and the model that bills you itself.

Do I need an OpenAI account to use it?

No. You use a Cheapest Imagen key, so there is no OpenAI account and no organization verification.

Does it work with the official OpenAI SDK?

Yes. In the Python and Node SDKs you change base_url and the key. Masks, streaming and transparent backgrounds are not supported.

What happens to my prompts and images?

The prompt goes to the model unchanged. An image is stored for 72 hours only if you ask for a link with response_format "url"; otherwise it is returned in the response and not stored.

Make your first image now

5 free images when you sign up, then from $0.008 an image.

Generate an image

Sign in

New accounts get 5 free images.