Context Window
The Whiteboard in the Meeting Room
5 min read
The Whiteboard in the Meeting Room
Imagine a meeting room with a whiteboard — everything on the whiteboard is visible and usable during the meeting, but once erased, it's gone.
The AI's context window is that whiteboard. Everything you've typed in the conversation — every message, every document you pasted — fits on this whiteboard. When it fills up, old content falls off. The larger the whiteboard, the longer and more complex conversations the AI can handle.
In Plain English
The context window is the maximum amount of text an AI model can 'see' and process at one time. It includes everything — your messages, the AI's replies, and any documents you share. When you exceed this limit, the AI forgets the oldest parts of the conversation.
The Technical Picture
The context window (or context length) defines the maximum number of tokens the model can attend to in a single forward pass. Both input (prompt + history) and output tokens count toward this limit. Longer context windows enable document-level tasks but increase computational cost quadratically for full attention.
Real-World Examples
- Claude has a 200K token context window — roughly a 500-page novel
- GPT-4 Turbo supports 128K tokens
- Gemini 1.5 Pro supports 1 million tokens — about 700,000 words
The context window is the AI's working memory — everything it can see at once in a conversation.