Four rates per model, and what each one charges

Input, cache read, cache write and output each have their own price. How one request is billed, step by step, and why the result is rounded up once.

Cent AIPublished 2 min read

Every model on Cent AI has four prices per million tokens: input, cache read, cache write and output. This post walks through what each one charges and how they add up to the cost of a single request.

The four rates

RateWhat it charges for
InputPrompt tokens the model reads fresh
Cache readPrompt tokens served from the provider's prompt cache
Cache writePrompt tokens stored in the cache for the next request
OutputTokens the model writes, reasoning tokens included

Cache read is the cheap one: on Cent AI it is 10% of the model's input rate. Cache write is billed at the input rate, for both the five-minute and the one-hour cache lifetime.

How one request is billed

The usage block of a response tells us how many tokens went where. A response in the OpenAI format looks like this:

{
  "prompt_tokens": 12000,
  "prompt_tokens_details": { "cached_tokens": 10000 },
  "completion_tokens": 800
}

Here 10,000 of the 12,000 prompt tokens came from the cache, so only 2,000 are fresh input. The cost is:

  1. Fresh input: 2,000 tokens at the input rate.
  2. Cache read: 10,000 tokens at the cache read rate.
  3. Output: 800 tokens at the output rate.

The three parts are added in nano-USD, a billionth of a dollar, and the sum is rounded up once for the whole request. Rounding once keeps the bill exact: nobody pays for rounding three times.

The minimum charge

By default, a successful request costs at least $0.00002. Very short requests, a one-word reply to a one-line prompt, can come out below that, and the minimum covers the fixed work every request takes. Failed requests that never reached the model cost nothing.

Try it with real prices

The calculator below uses the same price list as the pricing page. Pick a model and change the mix to see how cache changes the cost.

Cost per request

90% is an assumption you can change, not a promise. Cache is billed only from cache tokens the model provider reports.

See how many such requests it covers.

Estimate

deepseek-v4.1-flash
Fresh input99,000 tok
Cache read891,000 tok
Output10,000 tok

Per request$0.003956
At official pricesAt official prices $0.0235

Total$0.003956

An estimate at price version 4. Your bill comes from the usage the provider reports for each request.

Save $0.0195

Where to see your own numbers

Each response carries usage.cost, the amount charged for that request in USD. The dashboard lists the cost of every request, so you can check any line against the rates on the pricing page.

Sources

  1. cent-ai.dev/docs/concepts/request-cost (opens in a new tab) · accessed 2026-10-27
  2. cent-ai.dev/pricing (opens in a new tab) · accessed 2026-10-27
ShareXThreads