Perhaps. I found it was far too easy to hit the API limit with their old codex models, though that may have been limited to a small GPU cluster given it was pretty obscure compared to chatgpt and even davinci.