AI Glossary / Context window

⚙️ Technical

The context window is the maximum amount of text (measured in tokens) that the model "sees" simultaneously: your prompt, pasted materials, the conversation so far, and its own response. What doesn't fit, the model simply doesn't know - hence "forgetting" the beginning of long conversations. Modern windows reach hundreds of thousands, and in some models even a million tokens, allowing you to drop in an entire book or code repository. In practice, however, model performance can be uneven on very long contexts - key information is best placed at the beginning or end of the material.