I’ve noticed that during certain periods of time, the model (GPT-5.2) has an extremely high hallucination rate. I spend most of that time wasting tokens and money trying to correct the model, usually having to give up. During other periods of time, the model performs almost flawlessly. These periods seem to be quite random, and now I have to depend on pure luck to get the performance I need and my work done! And this happens despite no problems being announced on the status page.
Is anyone else experiencing this issue?