i’m starting to really understand why those that can afford it are building local llm/ai servers now. after a while a bunch of this stuff really starts to add up on the cost front so bringing a bunch of that local and inhouse makes a big difference, i’m starting to route requests to different models
