What is a model's context window and why does it matter?

The context window is the total amount of text a model can take into account at once: your messages, attached files, the earlier conversation and the answer it is writing. It is measured in tokens, not words, and works like the model's working memory, separate from what it learned in training.

It matters for three practical reasons:

• Limits. What does not fit has to be cut or shortened. Current top models list about 1 million tokens: GPT-6 and GPT-6.1 1,050,000, Gemini 4 Argon 1,048,576, Claude Opus, Sonnet and Haiku 5.5 1,000,000 (OpenRouter catalogue).
• Cost. In the API you pay for every input token, so sending a long history with each request makes every request more expensive.
• Quality. More context is not automatically better. Anthropic's documentation notes that accuracy and recall degrade as the token count grows (it calls this "context rot"), so it pays to include only what the task needs.

Chat apps such as claude.ai can manage the window on their side, which is one reason very long conversations may lose track of early details.