Note though that that a MAX 395 has half the memory bandwidth of a M4 Max chip, and the memory bandwidth is going to be the biggest limiting factor, so you'll likely be getting around half the tokens/second with that Framework Desktop.
replyThere's a reason why it's cheaper than the Mac equivalent and it's not all because of Apple's premium pricing =)
replyBut it's still the easiest and cleanest way to get decent local AI speeds on a non-Mac.