A context window is the working memory of an AI model - how much text it can keep track of while generating a response.

It's measured in tokens (roughly 3/4 of a word each). Claude 3.5 Sonnet has a 200K token window (about 150,000 words). GPT-4 ranges from 8K to 128K depending on the version.

When you hit the limit, the model starts forgetting earlier parts of your conversation or cuts off mid-response. Builders use larger windows for analyzing full codebases, processing long documents, or maintaining context across complex multi-step tasks.

Longer contexts cost more because the model does exponentially more computation. A 100K token input costs roughly 100x more to process than a 1K token input.