Ctrl+Enter to process
10
Context Windows
The context window is like the model's working memory — it can only hold and pay attention to a certain number of tokens at once. Everything it has seen in the current conversation (your messages, its responses, and any system instructions) must fit within this limit. Older parts of a very long conversation may get cut off if they don't fit.
Common misconception: A larger context window does not mean the model remembers everything equally well. Research shows that models tend to pay more attention to the beginning and end of their context, and can effectively "lose" information in the middle of a very long document. This is sometimes called the "lost in the middle" problem.
Messages
system
You are a helpful assistant.
user
What is a transformer model?