This is by far the best local llm model you can run on a 128GB Mac. It has only 6b active parameters making it super fast even on unified memory! And it's intelligent also :) Can't wait for them to launch a 27b/35b model on this architecture so we can run it on 36GB Macs(Mudit Gupta)