Comment by pixl97

In other industries of chance we have regulators that ensure compliance and that the providers aren't cheating.

At the end of the day the highest quality of benchmark tells you nothing if the man behind the curtain is constantly changing variables on you. You have no idea if you're really testing the same thing at all. So when you run your test at the top level on their system you're seeing lets say a 30% difference in quality most of the time, you have no idea if you should really only see a 5% difference in quality if you were running a local model with stable settings.