Calculator

What do routine prompts cost you?

Most prompts are routine: a rename, a small fix, a test, a commit. Sent on a frontier model, they cost frontier prices, in Cursor, Claude Code or Codex alike. Move the sliders to your own numbers. It all runs in your browser.

Your tool

The model you usually have selectedPrompts a day60Working days a month21Share that is routine (rename, small fix, test, commit)50%Where routine prompts could run

Tokens per prompt (advanced)

Input tokens (context) per prompt20kShare of input read from the prompt cache70%Output tokens per prompt800

Estimated model spend a month

$71.82

With routine prompts on the cheaper model

$50.27

About

$21.55

a month not spent, before Pro’s $6.99 (about $14.56 ahead).

An estimate from your inputs and per-token list prices (Cursor’s published model list, 29 September 2026); not a quote. Plan allowances, Max mode and cache writes are not included. During the trial, your console shows the real saving from your own prompts.

Start free trial

Pro is free for 30 days, no card. Your console then shows every switch, and in Cursor what it actually saved you.

Prices are list prices per million tokens; the source for each tool is under the result. On a Claude or ChatGPT subscription you pay a flat price instead of per token: there, the same share is how much of your plan’s usage routine prompts take up.