How much text a model can consider at once, measured in tokens. Long context is why you can paste a full chapter and ask questions about it, until you exceed the window and the model quietly forgets the start.