Tokvera

Calculator

AI bill leak calculator

How much of your LLM bill goes on input your requests repeat and nothing caches? Enter your own numbers: there are no defaults and no guessed prices.

Your numbers

Calls to the model API in a month.

Average prompt size, including the system prompt and context.

Average completion size.

From your provider’s price list, in any currency.

In the same currency.

How much of each prompt is the same as in other requests and is billed at the full input price.

Estimate

Monthly spend
–
Monthly spend on repeated, uncached input
–
Share of the total
–

An estimate computed only from your inputs, in the currency of your prices. Not a measurement.Run Tokvera to measure whether your repeated prompts are worth caching.

The calculator needs JavaScript. The formula below gives the same numbers by hand.

How it is calculated

With R requests a month, Tin and Touttokens per request, prices Pin and Pout per million tokens, and a repeated, uncached share h of the input:

  • Monthly spend = R × (Tin × Pin +Tout × Pout) ÷ 1,000,000
  • Spend on repeated, uncached input = R × Tin × h ×Pin ÷ 1,000,000

The second line is what the repeated part of your prompts costs at the full input price. It is not a forecast of what caching would change: that depends on your provider’s cache pricing and on how often requests really repeat, which only a measurement shows.

From an estimate to a measurement

Tokvera measures your endpoint instead of estimating it: Time to first token and latency percentiles at rising concurrency, error rates, cost per 1,000 requests, and whether repeated prompts are worth caching.

The AI Performance & Observability Audit instruments your production traffic and leaves a ranked fix list.

From $3,500, about 2 weeks.

Book a callPricing