Context window
A context window is the maximum amount of text or conversational history that an AI model can process and reference at one time when generating a response. It is typically measured in tokens, which represent word fragments rather than whole words, and its size determines how much prior conversation, instructions, or uploaded text the model can take into account before it begins to lose access to earlier information. Once a conversation exceeds this limit, older content is generally truncated, summarized, or dropped from the model's active reference.
For users of conversational AI platforms, the context window size has a direct effect on continuity and coherence during longer interactions. A larger context window allows extended conversations, detailed scenario-building, or ongoing character interactions to remain consistent over time, while a smaller one may cause the system to forget earlier details, contradict previous statements, or require users to repeat information periodically. This is distinct from persistence between separate sessions, since a context window concerns memory within a single active session or a single continuous exchange, though some platforms combine both mechanisms.
Context window size is generally disclosed, if at all, in technical specifications rather than marketing material, and it can vary between subscription tiers, with paid plans sometimes offering larger windows than free access. Understanding this limit helps users anticipate when a system might lose track of earlier context and plan accordingly, particularly for interactions involving extended narratives or detailed ongoing instructions.
Updated: