AIFastAIFast Developer Docs
TOKEN COST LAB

LLM API Token Cost Calculator

Estimate the cost of single and batch calls based on the current price, each task Token, number of tasks and failure retry ratio.

The price is filled in by you according to the current page Display retry costs separately Data is only calculated in the browser
01 · Enter calculation conditions

Fill in the current model price and usage

Model price, Token and task quantity are not sent to the server.

02 · Cost results

Waiting for calculation

Fill in the current model price and real business usage, and the results will display the basic cost, retry cost, and total cost respectively.

View cost control methods
Compute boundaries

Don’t just compare price per million Tokens

Lower unit prices do not necessarily translate into lower complete mission costs. Retries on failure, long contexts, tool loops, and manual rework can all change the final outcome.

  1. 01

    Use logs to count real Tokens

    response from usage and sample from the bill, don’t just guess based on the number of characters in the prompt word.

  2. 02

    Count failed requests into account

    Counts additional requests resulting from timeouts, 429s, 5xx, format errors, and Agent retries.

  3. 03

    Compare by complete tasks

    At the same time, record the completion rate, delay, rework time and token cost, and then decide the model and routing.

Costing FAQ

Results are estimates based on input conditions and do not replace console billing and production logs.

Why not just build in the price of each model?

Model prices, billing units, caches, and channel discounts vary. It is filled in by the user according to the current console or price page, which makes it easier to retain the calculation basis and avoid wrong decisions caused by old quotations.

How should I fill in the failed retry ratio?

Use the real request log to count the number of additional requests divided by the number of scheduled tasks. For example, if 100 tasks are scheduled and 8 additional retries are performed, 8% can be filled. If there is no data, fill in 0 first, and then update it with the log after going online.

Is the Token cost equal to the complete business cost?

Not equal to. The full cost may also include images, video, retrieval, caching, tool calls, networking, manual rework, and failure recovery. This tool only estimates the filled-in text token and retry parts.

Will the calculation results be uploaded?

No. All input and calculations are completed in the current browser, and model prices, tokens, or task quantities will not be submitted.

Model prices and availability start withLive price pageSubject to the console. Continue readingAPI cost estimation and usage controlandModel API comparison test template。