Context Window

The context window is the maximum amount of text — measured in tokens — that a language model can process as input and store in its short-term memory during a single interaction. This limit determines how much history, documentation, or code can be offered to the model's 'attention' at once for coherent responses. The increase in context window size (e.g., 128k or 1M tokens) enables the analysis of entire books, legal contracts, or complex codebases, significantly extending the practical applicability of LLMs in information retrieval and summarization.