Knowledge
What is a token?
A token is a piece of text a language model works in; it sets how much text fits and what it costs.

A token is a piece of text, roughly part of a word. A language model processes text in tokens.
What you read about it
- Google reported that Gemini 1.5 Pro handles up to 1 million tokens stably. With that much context you can give a whole file of documents to the model at once: Gemini 1.5 Pro.
- Claude 3.5 Sonnet reads up to 200,000 tokens at once. Through the API it costs $3 per million input tokens and $15 per million output tokens: Claude 3.5 Sonnet.
- The more tokens in the context window, the worse the model is at finding information in it: from prompt to context engineering.
- A model that thinks for a long time costs more time and tokens: QwQ.

Why it matters to you
Tokens set how much text a model reads at once and what a call costs. More context is no guarantee that the model reads everything well. So pass only what the task needs. Test with a question whose answer you already know, deep inside your own document.

Written on 13 November 2026, based on the articles and sources below.