Token Cost Calculator
Inference is priced per token, billed per million, and paid at volume — which makes a feature that feels cheap per request expensive per year. Put in your token counts and your provider's prices; the yearly number is usually the one that changes the decision.
Use your provider's current price sheet — prices move; this calculator never assumes them for you.
Per request
$0.0135
Per day
$135.00
Per month (30d)
$4,050.00
Per year
$49,275.00
Where the money goes
Input tokens · 44%
Output tokens · 56%
Output tokens usually cost several times more per token — trimming verbose responses often saves more than switching models.
How to read it: the input/output split matters more than most model comparisons — output tokens typically cost several times more per token, so a verbose system prompt is cheap but a verbose answer is not. Cutting response length, caching repeated context, and routing easy requests to smaller models usually move this number more than renegotiating a price sheet.
Embed this tool
Paste this snippet into any page to run the calculator there.