For day to day LLM experimentation (and even some business use cases), I'd say Apple Silicon would be first choice for me.