Glossary
Context window
The context window is how much text an AI can keep in mind at once: your messages, its replies, and any pasted documents, measured in tokens. When a chat outgrows it, the AI starts forgetting the oldest parts.
The context window is the AI’s working desk. Everything in the current conversation (your questions, its answers, the PDF you pasted) has to fit on that desk, measured in tokens. When the desk is full, the oldest papers slide off the edge, quietly.
That’s why a long chat can get weird: you told the AI on Monday that your B&B has six rooms, and by message eighty it’s inventing a swimming pool. It didn’t lie out of malice; that detail simply fell off the desk.
Three habits fix most of it. Start a fresh chat per topic. Re-paste the key facts when a conversation runs long. And for big documents, check the model’s window size first. Modern models fit hundreds of pages, but not infinitely many.
Where you’ll meet this
Model comparison pages brag about it (“200K context,” “1M tokens”), and you’ll feel it as “this conversation is too long” warnings or answers that ignore your earlier instructions. Window size differs a lot between models and tiers, and it’s one of the things our AI chooser weighs when matching you to an assistant.