I seem to be getting 503 errors at >90% of my gemini-2.5-pro API requests, even with a tier-1 paid account. This makes it practically impossible to use the model even for prototyping. The task is relatively simple, single image interpretation, I don’t think I even exceed 2k tokens per request. I have to use pro, I can’t just switch to flash (where that’s less of an issue). I can understand the occasional server-side error, but this makes the service pretty much unusable.
Does anyone else have the same error rate? I don’t remember it being this bad a couple months back. Pretty frustrating.
