Context window
Also called context length.
A contextContext is the information available to an AI model for a particular response, including instructions, conversation, and supplied material. window is how much content a model can handle at once, measured in tokens, with limits that depend on the model.
Example
You paste a 200-page manual into a chat that only keeps a smaller window. The model may answer from the portion that fit, not from the page you assume it still sees.
Why it matters
Input, generated output, and sometimes other internal tokens share that budget. A long chat is not proof that every earlier message is still available. When the window is full, something is left out or summarized.
Common confusion
A long conversation does not mean the model still has the first message. The window is a limit, not a filing cabinet.
Sources
Updated September 30, 2026.