Hi everyone and dear @[email protected], hope you're all doing well.

I need to vent a bit and ask a question: Why do text model constantly change their behavior throughout the day? This has been a persistent issue, and honestly, it’s getting super frustrating. I have no idea how we're supposed to find a reliable workaround when the output quality keeps shifting like this.

So far, I’ve been through 3 different text model updates, and this issue was present in all of them. The AI performs great during certain hours of the day, but then its performance just plummets during others (one specific time slot is an absolute disaster). By the way, I’ve heard that people experience the exact same issue with image generation models too, so it seems like a widespread problem.

This exact thing just happened with the latest text model update. During some hours, it’s like its vocabulary suddenly shrinks, the reasoning goes out the window, and it just keeps repeating the exact same words over and over.

What is the actual technical reason behind this? It’s incredibly annoying.