The context window is the working space of an AI model. It is the total number of tokens the model can see at once: instructions, conversation so far, any documents you supply and the answer it is writing. Anything outside the window is simply not available to the model.
Window sizes differ between models and have grown considerably, with some now able to hold the equivalent of several books. When a conversation or document exceeds the limit, the system must cut, summarise or drop older content, and the model can then appear to forget earlier details.
Hospitality and retail operators run into this when they paste long supplier contracts, price lists or a year of customer reviews into a tool, or when a customer chat runs for a long time. A bigger window helps, but bigger is not free: more tokens in means higher cost and slower replies.
Even within the limit, models can pay less attention to information buried in the middle of very long inputs. Do not assume that fitting something in the window means it will be used well. For large knowledge bases, retrieval is usually a better approach than pasting everything in.
Knowing the window helps you design assistants that stay accurate through long chats and keeps costs predictable.