Whichever one tell you exactly what you want to hear?

text-davinci-003 model. It has been deprecated by OpenAI but it is still available in Microsoft Azure OpenAI service.