Before a model can think about your sentence, it shatters it into small chunks called tokens. Tokens are how models read, how they write — and how every bill is counted.
Each block below is one token. Notice what happens: common words stay whole, long or unusual words get chopped into pieces, and spaces ride along inside the blocks. Try the presets, or write your own.
␣ marks a space living inside a token. Emoji and symbols often cost 2+ tokens (look for the ×2).
This is a teaching simulator using a simple rule. Real models use a learned tokenizer, so exact splits vary — but the patterns you see here are real.
For everyday English, one token lands at roughly 4 characters, or about ¾ of a word. So 1,000 tokens ≈ 750 words ≈ a page and a half.
cat1 tokenlanguage1 tokenantidisestablishmentarianism7 tokens🎮2 tokensYou pay for tokens going in (your prompt + everything the model can see) and tokens coming out (the reply). Output is the pricey side — usually 5× the input rate. Pick a model, dial the size of the job, and watch the cost move.
Models chop text into tokens. Common words are one token; long or rare words get split into pieces.
Spaces, punctuation, and even a single emoji all become tokens. Nothing is free.
A rough ruler for English: 1,000 tokens ≈ 750 words. Handy for guessing a bill before you press send.
You pay for the prompt in and the reply out — and output costs about 5× more. Cheaper model = smaller bill.