About degraded quality of gpt-3.5-turbo's responses

I’m using gpt-3.5-turbo for a specific task that involves a system prompt. Upto yesterday, it was working fine but today, it’s responses became worse and it is having trouble following the system prompt. I’m just wondering, has anyone else experienced this or not?

Yes, but I started to experience it since they came up with 4o. Neither 4o nor 3.5 can do what 3.5 used to. I had to switch my tasks to gpt-4 . It looks to me like there is no 3.5 anymore, it all goes to 4o