In the model settings, it would be nice to be able to adjust the maximum number of output tokens.