Beginner
What is a Context Window?
Why a model 'forgets', what 128K or 1M means, and how it affects cost.
AI HQ Editorial3 min read
The model's working memory
The context window is the maximum amount of text (in tokens) a model can consider at once: your instructions, everything pasted in, and the conversation so far, plus its reply. A 128K window is roughly 100,000 words; a 1M window is about eight novels.
When a conversation exceeds the window, older content is dropped or summarised. That is the 'forgetting' people notice in long chats.
Bigger is not free
Everything in the window is billed as input tokens on every turn. Long conversations and large pasted documents are the fastest way to run up a bill. Context length is listed for every model on our Models page.
