Apple’s big AI week ends with Mac + Muse AI supercomputers

Apple's machine learning team introduced a new version of MLX that enables running large language models locally on Mac Studio with Apple silicon.

Apple and Meta collaborated to make it possible to run Muse Spark Open on two or four Mac Studios, achieving over 36 tokens per second on a cluster of four Mac Studios.

The release includes software support for Neural Accelerators in the M5 family, low-latency communication via RDMA over Thunderbolt, and efficient implementations of mixture of experts architectures.