Sorry but this is a pretty delusional take. No, you cant run the same level or even close to that level of model on a mac, not even on a 512gb studio. You can run good models sure, but not models anywhere close to the capabilities of these ones yet, despite what idiots on Twitter keep spouting.

The "run locally" thing is a red herring. The fact that anyone can be an inference providers and offer them as a service is what we need to focus on. If they are "good enough" then market forces will stop Anthropic/OpenAI from charging monopoly rents.