What is a token in AI? Meaning and examples

A token is the small chunk of text — roughly a short word or part of a word — that language models read and generate, and that providers usually charge by.

In plain English

Language models do not deal in letters or whole words. They break text into tokens: common words are often a single token, rarer or longer ones are split into pieces, and punctuation and spaces count too. For everyday English, a rough rule of thumb is that 1,000 tokens is about 750 words, or roughly a page and a half.

The model turns these tokens into numbers, predicts which token should come next, and repeats — which is why answers appear word by word rather than all at once.

Why it matters

Tokens are the unit of measurement for almost everything practical about these systems. Usage is priced per token, speed is quoted in tokens per second, and limits on how much a model can read or write at one time are expressed in tokens. If you are budgeting for an AI feature at work, you are really budgeting tokens.

An example

Pasting a 20-page policy document into a chatbot and asking for a summary might consume around 10,000 tokens of input, while the summary itself uses a few hundred output tokens. Do that fifty times a day across a team and the token count — and the bill — adds up in a way a single chat never suggests.

What to watch out for

Tokenisation is why models can stumble on tasks that look trivial, such as counting the letters in a word or handling unusual spellings: they simply do not see individual characters. Languages other than English, code and numbers often use more tokens per unit of meaning, so the same text can cost noticeably more in one language than another.