GLOSSARY
Context Window
How much text a model can consider at once — its working memory, measured in tokens, spanning your conversation and any attached documents.
The context window is a model's working memory: everything — your messages, the assistant's replies, uploaded documents — must fit inside it. Small windows forget the beginning of long chats; large ones (Claude's 200K tokens, Kimi's millions of characters) can hold entire books or codebases.
A bigger window is not automatically better: models can lose focus in very long contexts, and cost scales with size. Tools work around limits with techniques like retrieval (RAG), which fetches only relevant passages instead of stuffing everything in.