What is Tokens?

In AI, tokens are small pieces of text that a model reads and writes, like parts of words, whole words, or punctuation. Models do not “see” sentences the way people do. They break text into tokens and predict the next token. For everyday users, tokens matter because they affect how much text the AI can handle at once and can influence cost or limits in some apps. A downside is that long conversations can get cut off or older details may be forgotten when token limits are reached.

Tokens are little chunks of writing that the computer uses, kind of like building blocks for words and sentences. Instead of reading a whole sentence at once, it reads it in small pieces. Then it tries to guess what piece should come next.

Think of a long sentence made from letter tiles or word tiles. The AI picks up a few tiles at a time and tries to choose the next tile that makes the most sense. If you give it too many tiles at once, it may have to put some aside, and that can make it forget earlier parts.