Context window
The amount of text an AI model can take into account at one time, including your messages, any documents you share and its own replies.
Think of the context window as the model's working desk. Everything on the desk can be used to form an answer: the conversation so far, a pasted report, your instructions. Anything that does not fit on the desk is out of view. The size is measured in tokens.
If you paste a very long contract into a chatbot and ask about a clause near the end, a model with a small context window may not be able to see that part. In a long conversation, the earliest messages can fall off the desk, and the tool may seem to forget what you said at the start.
A bigger window lets a model handle longer documents, but it does not guarantee it pays equal attention to everything inside. Details buried in the middle of a long text can still be missed, so ask about specific sections and check the answers.
Related: Token, Large language model (LLM), Prompt