Context Window

Definition: The maximum amount of text — measured in tokens, not words — that a model can hold in view at once, including everything you've said in the conversation so far, any documents you've pasted in, and the reply it's generating.

Why it matters: Once a conversation or document exceeds the context window, the model doesn't remember the earliest parts anymore — it has genuinely lost them, not just deprioritized them. A model that seems to "forget" something you said earlier in a long chat has usually just run past its window.

Example: A 200,000-token context window fits roughly a 500-page book — but a long back-and-forth conversation with pasted documents can fill that up faster than you'd expect.