The maximum number of tokens a model can consider at once.
The context window limits the combined input and output tokens for a request or conversation. Models range from tens of thousands to millions of tokens.
A larger window lets you send more documents without chunking, but it also raises the maximum possible input bill and can increase latency.