Until last two days, I have had ability to get consistent answers through the same prompt engineering, same token count. I made around 5000-10000 calls to be able to call it that way. Literally with two eyes and analyzed numbers, it is very noticeable that last two days it’s lowered in intelligence. I’m wondering is it due to new update with fine tuning and open ai team included chat 3,5 (perhaps turbo too) in there? Even the response time is lower.
