Cheapest AI image API: what you actually pay per image
Image pricing has nothing in common with token pricing. There is no input/output split, no context window, no cache discount — you pay for a picture, and the two numbers that decide your bill are resolution and how the provider rounds. This page shows real per-image prices from our live catalogue, including the models where we are not the cheapest.
Three billing shapes you will meet
- Per image (flat). Easiest to budget. The catch: a "per image" price usually means one resolution tier — 4K costs more or is silently downscaled.
- Per megapixel. Common with Flux-family models. Rounding matters: a 1024×1024 output is 1.05 MP, which vendors round up to 2 MP. Your effective price can be double the headline.
- Per second of compute. Used by the fast inference providers. Same model, same image, different price depending on how busy their GPUs are.
When you compare two providers, ask for the price of one specific output size. A "$0.003 per image" headline is usually a 512px quick-model result, not the 1024px image your product ships.
Live per-image prices from our catalogue
Read directly from the catalogue, so it cannot go stale. Prices are USD per generated image, sorted cheapest first.
| Model | Vendor | Per image | List price | vs list |
|---|---|---|---|---|
| — | — | — | — | — |
Where we are not the cheapest
Being straight about this is worth more than a fake discount. Some image models on our catalogue cost more than the vendor's own list price, because they route through a higher-fidelity channel that produces fewer retries. gpt-image-2 is the clearest example. If your workload is retry-sensitive — product photography, text-in-image, brand assets — the extra cents per image are usually cheaper than regenerating failures.
The table above computes the delta against list price for every model, so you can see exactly which ones carry a premium and which ones you should route to. We do not hide the expensive rows behind an average.
How to generate your first image
OpenAI-compatible image endpoint. Same shape as the official API:
POST https://tapixai.com/v1/images/generations
{"model":"az-gpt-image-2.5-flare","prompt":"a isometric server rack at dusk","n":1}
No subscription, no seat. Every generated image is deducted from your prepaid balance and shows up in the console with its exact per-image cost.
How to decide, in one paragraph
If you ship more than a few hundred images a month, per-image APIs beat flat subscriptions almost every time, because you stop paying for capacity you do not use. Below that, a $10 top-up on a per-image gateway is still cheaper than one month of most image SaaS. The only case where a subscription wins is a fixed, high, predictable volume on a single model — and even then, price the same output size before you commit.
Try it free with $1 credit See full pricing