The maximum amount of text (measured in tokens) that an LLM can read and reason over in a single query.
The Context Window acts as the model's working memory for a single request. Larger context windows let organizations place long documents, sets of contracts, or large parts of a codebase in one prompt. However, filling the window toward its limit can weaken how well the model attends to relevant details, and inference compute rises steeply as inputs get longer.
See also Private AI, Digital Transformation and Enterprise Architecture.