Any guesses on using this for LLM inference how it compares performance-wise (like tokens/sec ultimately) to like the new mac mini m4pro? I'm not sure how to fi

Any guesses on using this for LLM inference how it compares performance-wise (like tokens/sec ultimately) to like the new mac mini m4pro? I'm not sure how to figure that out.