3.3 What Are Context and the Context Window?

The current information a model can refer to while answering is called context. It may include your present question, earlier conversation, uploaded material, product instructions, and recent tool results.

Imagine context as the AI's current worktable. Information on the table can participate in the task. A private fact that was never placed there is not automatically known.

A context window has a capacity

The context window is the amount of context a model can process at once. It is normally measured in tokens, not pages. Models and products have different capacities, and a product may shorten or summarize long material.

When a conversation becomes very long, earlier information may no longer be used completely. The AI may forget that you said “no peanuts” at the beginning or continue following an old goal that has changed. It is not deliberately disobeying; the key information may no longer be available in the current content.

Keep critical constraints near the task, summarize a long conversation periodically, and provide essential background when starting a new chat. More context is not always better: relevant, explicit, non-conflicting information is more useful than a disorderly pile.

Chat history, product “memory,” and context are not synonyms. History may be saved without every old message appearing in current context. Whether memory exists and what it stores depend on product settings.