Apple's Mac Studio with the M4 Max offers a compelling local AI platform due to its large unified memory pools and high memory bandwidth, enabling faster tokens-per-second throughput than Nvidia's GB10 or AMD's Strix Halo. The M4 Max version of the Mac Studio, last refreshed in March 2025, includes two variants: a base model with a 14-core CPU and 410GB/s of memory bandwidth, and an upgraded 16-core CPU version with 546GB/s of memory bandwidth. Apple lent a 16-core CPU, 40-core GPU, and 128GB memory M4 Max system for LLM-specific tests, priced at $3,699. The decode phase of LLM inference is sequential, making memory bandwidth critical for streaming model weights to the GPU. Apple Silicon is the only platform offering both large memory pools and high memory bandwidth at a relatively reasonable cost, as no other unified memory SoC matches its wide memory bus.