上下文窗口

The maximum number of tokens a model can process in one request.

The context window decides how much content a model can see at once. Larger windows enable long documents, long conversations and big codebases, but they raise cost and attention efficiency concerns.