Can the option to turn on/off deep thinking on the model be added, this is important