Glossary · Term

Context Window

Also known as: context window

A context window is the amount of content an AI model can remember and refer to within a single conversation.

Context refers to the range of content that AI can remember and refer to in one conversation, and that limit is called the context window. If you compare it to the space on your desk, you can refer to the materials on the desk right away, but when the space is full, you have to get rid of old materials first.

When you give a document to AI and ask it to summarize or continue a long conversation, only the content within this range is actually reflected. Therefore, how long the models can handle the context has become a key indicator in the performance competition, and models that can fit an entire book have emerged.

This limitation is often the reason why AI answers in long conversations as if it has forgotten the first part. In addition, even though the context is long, not all of the content within it is used evenly, so it has been reported that information in the middle is missed.

✅ Why it matters

⚠️ Limits and debates

← View all glossary entries