Estimate your monthly AI API costs across multiple models. Built for developers who want control over their AI infrastructure.
Data stays in your browser. No server storage.
A hypothetical model uses two thousand tokens per request and ten thousand monthly requests. Their product is twenty million tokens. Dividing by one thousand makes twenty thousand rate units. At an entered two cents per thousand, the monthly estimate is four hundred dollars. These dollar bars use one pixel per dollar. Rates and traffic are user assumptions, not current provider prices, actual usage or a bill. Keeping the same entered rate and tokens per request, twenty thousand requests instead of ten thousand doubles twenty million tokens to forty million and four hundred dollars to eight hundred. The common zero and half pixel per dollar bars show two hundred and four hundred pixels. The source has one rate, with no separate input/output, caching, taxes, retry or provider billing rules. Hypothetical four hundred and one hundred dollar rows sum to five hundred dollars. Common one pixel per dollar teaching bars show those amounts; the source breakdown instead divides each row by the maximum row, yielding one hundred and twenty five percent before browser flex sizing. It is not a normalized eighty and twenty percent share chart. Remove deletes a row and the text summary follows the remaining rows. Reload loses this in-memory list; no server or account data is connected.