I have been creating completions with the model, text-davinci-003 for one of my projects. When the ChatGPT API was widely released, I noticed my app disregarded the best_of value and gave bland answers.
The documentation states that completions should still accept best_of values. However, I wrote a basic API Completion call and it works - up until you add in the best_of parameter. Is this a bug, or does the documentation need updating?