If you can run sota on a 40k setup, why do openai etc spend maybe 100x that?
Obvious one: Because they are serving it to millions of people at the same time, not just one local user
Obvious one: Because they are serving it to millions of people at the same time, not just one local user