Paste any text to estimate how many tokens it is, then see what it would cost to send as a prompt across the major models. Everything runs in your browser — nothing you paste is uploaded.
Token count is an estimate (~4 characters per token), usually within 10–20% of a model's real tokenizer. Prices in USD, reviewed —.
Your text as input + expected output. Cheapest first.
Roughly four characters a token
For English prose, one token ≈ 4 characters ≈ 0.75 words. Code, other languages, and symbol-heavy text break that ratio, so treat this as a close estimate, not a hard count.
Watch the context window
Every model caps total tokens (prompt + output). If your document is bigger than the window, you'll need to trim, chunk, or pick a longer-context model — better to know before the call fails.
System prompts add up
A long system prompt is sent on every call. Paste it here to see its per-call cost — trimming a verbose system prompt is an easy, repeatable saving at scale.
It uses the standard ~4-characters-per-token rule, usually within 10–20% of a real tokenizer. Code, non-English text, and heavy symbols tokenize differently. Fine for budgeting; for hard limits, use the provider's official tokenizer.
You're billed per token, and each model caps total tokens (prompt + output). Token count drives both your cost and whether a prompt fits the context window.
Output is usually priced higher — often several times — because generation is more compute-intensive. This tool prices your text as input and adds your expected output for the full call cost.
No. Counting and pricing happen entirely in your browser. What you paste never leaves your device.
Get an email when we ship the next AI cost or prompt tool.
No spam, no signup needed to use any tool. Unsubscribe any time.