Comment by verdverm

Interesting, so you are saying you have an easier time generating JSON using the single prompt models like text-davinci and text-bison, rather than the chat versions?