Will there be any surprises when the same prompt words are used with different models?