Here's what they mention:
> Because gpt-3.5-turbo performs at a similar capability to text-davinci-003 but at 10% the price per token, we recommend gpt-3.5-turbo for most use cases.
Is capability to support trained models only additional advantage of da-vinci-003?
