Represents a candidate response from the model.
The content of the candidate.
The reason why the model stopped generating tokens.
Safety ratings for the content.