Token
A token is the unit AI models actually read and write — a word fragment of a few characters, with roughly 750 English words fitting into 1,000 tokens.
Models never see letters or whole words. Before any text reaches one, it is chopped into tokens from a fixed vocabulary: common words survive as one piece, rarer words are split into fragments, and code or unusual spellings splinter into many. Everything about a model is priced and measured in this unit — API rates per million tokens, context windows in thousands of tokens, generation speed in tokens per second.
The fragment approach is what lets a fixed vocabulary of some tens of thousands of pieces cover any input at all — every language, every typo, every programming language. A side effect worth knowing: languages other than English usually need more tokens for the same sentence, so the same conversation can cost more and fill the context faster.
For everyday intuition, two rules of thumb carry far: a token is about four characters of English, and a thousand tokens is a longish e-mail.
Where you see it on the radar
The cost calculator states its workloads in tokens per day — the unit every price on the page is measured in, from API rates to what your own hardware effectively costs.
See what a day of tokens costs →Related terms