# Token

> A chunk of text the AI reads and counts, roughly three-quarters of a word. Tokens drive limits, cost, and speed.

**Category:** Under the hood

**A token is a small piece of text.** AI does not read whole words; it breaks text into tokens, each roughly four characters or three-quarters of a word. A long word can split into several tokens, and a page is about 500.

Tokens are the unit working behind the scenes: **the [context window](/glossary#context-window), speed, and cost are all measured in them.** You rarely count tokens by hand, but knowing they exist explains why very long chats and huge files start to strain.

## At work

Uploading a 60-page PDF when you only need two pages burns tokens fast and fills the window. Paste the part that matters instead.

## Often confused with

A word. Most short words are one token, but long or unusual words split into several, and spaces and punctuation count too.

## Related terms

- Context window
- Context

## See it in a lesson

[Context and tokens](/en/learn/working-smart/context-and-tokens)

## Further reading

- [Context windows (Anthropic docs)](https://docs.claude.com/en/docs/build-with-claude/context-windows): How tokens fill the window.
- [How do usage and length limits work? (Help)](https://support.claude.com/en/articles/11647753-how-do-usage-and-length-limits-work): How token counts turn into limits.
- [Large language model (Wikipedia)](https://en.wikipedia.org/wiki/Large_language_model): Where tokens come from in how models read text.

---

Last updated: 24/07/2026  
Source: Rollo Academy glossary