Not sure if anyone else has seen this but I’ve noticed a distinct difference in the number of rejected responses using the Azure version of GPT and yet when I run the same query through the GPT moderation tool, nothing is flagged and it is answered fine by both the API and ChatGPT.
Perhaps microsoft is trying out a new filtering methodology but wanted to mention it in case anyone, like me, is considering having both setup and failing over in case of an outage.
Also worth noting if people are using both that the embedding’s in Azure can only take one input at a time but the error message says they are working on it.