Expand description
Given a prompt, the model will return one or more predicted completions,
and can also return the probabilities of alternative tokens at each position.
Compared to the chat API, this one does not provide the ability to have
multiple rounds of conversation. This API is getting deprecated in favor of the
chat API.