I mean it's not hard to understand that if good model can run on consumer hardware, even better models can run in data centers