Maybe it is related to them updating the backend to support GPT-5 in production. I appreciate these downtimes should be planned, but since it's OpenAI, this may be it. One can only hope, though.