Context window

The context window is how much text an LLM can “see” at once—both your prompt and the response it generates. Think of it as the AI’s short-term memory. When the conversation exceeds this limit, older parts are forgotten or compressed.


In one sentence

The context window is the AI’s working memory: how much text it can keep in mind at once.


What it's for

If you paste a 50-page report into Claude and it starts missing details from page 1, you’ve hit the context window limit. For most users this only matters with very long documents or multi-hour conversations.


How to think about it

Imagine a whiteboard that can only hold a few paragraphs. Write more, and the oldest words get erased. That’s the context window. Claude (200K tokens) ~= 150,000 words. GPT-4 (128K tokens) ~= 96,000 words.


Common misunderstandings

People think context window = intelligence. It is not. A model with a small window can be smarter than one with a huge window. The window is just memory size, not reasoning ability.


If you want to try it

Open Claude.ai. Paste a long article. Ask it to summarize the first paragraph. Then keep chatting for 20+ exchanges. Ask about the first paragraph again. See if it remembers accurately.


Sources