Context window

The context window is how much text a model can “see” at one time — your message, the conversation so far, and everything you fed it — all counted in tokens.

Think of it as the model’s short-term memory. Whatever doesn’t fit in the window simply isn’t there: no memory of it, no way to reference it.

That’s why:

• very long conversations can start to “forget” the beginning,
• pasting a huge document can push out earlier messages,
• tools like RAG exist — they re-insert the most relevant pages of a big document right when the model needs them.