Has anyone else experienced drop in realtime api’s performance recently. Model is now not able to answer questions based on workflow which it didn’t have problem answering couple days ago. Same prompt, same workflow, same code but worse performance
1 Like
Hello!
I saw your post about the drop in API response performance, and I wanted to share my perspective and a possible hypothesis that might help.
In my own experiments, I actually use intentionally ambiguous prompts to put a load on the API and observe a decrease in performance.
From this perspective, it’s possible that even with the same prompts, there may still be subtle variations or ambiguities affecting the results.
If you could make your questions and workflow even more transparent and clarify your instructions, it might help improve the API’s response speed.
I hope this perspective is helpful to you!