01 / The Model
Token
The unit a model reads and writes, roughly a word and often less. It matters because everything you pay for, and every limit you run into, is counted in tokens rather than in pages or minutes.
How you’ll hear it
“Why did the bill jump this month?”
“We sent three times as many tokens. The documents we attach count every time.”
Avoid
Equating tokens with words. They are usually smaller, and the ratio changes by language.
The longer version
Models read and write in tokens, fragments of text often shorter than a word. Pricing, speed and the limit on what a model can see are all measured in them. Languages other than English, Arabic and Finnish among them, often need more tokens for the same meaning, which shows up in cost.