>> Ignore anyone using an API (even for the same model) they are most likely redirected and using an extremely similar, but separate set of resources. (possibly even using a higher load in the off hours than during normal usage.) Or any other type of automated (non human) input.
For example: Weekdays 2:30 - 5:30 am EST. Sundays +2 hours<p>I noticed with the original Open AI releases though the magnitude decreased over time.<p>>> I have been using Gemini 1.5 pro (experimental - 0801 / 2,097,152 token count) since Friday and for the first time - Shockingly good but only during the off hours? I tested both my account and my wife's account with similar queries and was less impressed during peak hours.
Yes, this happens, there's happening some throttling, I've seen questions like this one regarding the same issue across several LLM providers ("works faster, better, solves better at night").