Tokens and the Context Window

A model reads tokens, which are small pieces of text. About three quarters of an English word is one token. The context window is the most tokens a model can use at once. Cost follows tokens in and out, so keep prompts short and focused.

Token

A small piece of text that a model reads and writes.

Context window

The most tokens a model can use at once.

Input and output

Tokens sent in, and tokens written back.