HACKER Q&A
📣 doctoboggan

Tips for using OpenAI’s lower end models


I’ve struggled to get good results from openai’s lower cost models like Curie, Babbage, and Ada.

Does anyone have tips or tricks to squeeze the most performance out of these? Or specific tasks that they are able to excel at?


  👤 cloudking Accepted Answer ✓

👤 gardenhedge
These models generally sucked until recently which is why there is so much craze about ChatGPT. They finally got it (somewhat) right.

👤 speedgoose
I found them pretty bad even when fine tuned for a simple classification task. I decided to ignore them, Davinci is too much better.