Glossary

Token

Short answer

The smallest unit a model breaks a prompt into; it defines the prompt length limit.

Models read a prompt token by token, not word by word. A token usually corresponds to a word or part of one.

Every model has a token limit, and the end of a prompt that exceeds it is usually silently ignored. That is the technical explanation for why very long prompts do not work.

The practical consequence: put the most important information at the start of the prompt. Details left at the end are more likely to be disregarded.