Apple's M5 Ultra Chip Demonstrates Strong AI Capabilities for Local LLM Workloads

Six weeks after its introduction, Apple's M5 Ultra chip is proving its significant impact on AI workloads, particularly for running large language models (LLMs) locally on devices. The updated Mac Studio, powered by the M5 Ultra, features a 36-core CPU, an 80-core GPU with dedicated Neural Accelerators for the first time on an Ultra-class chip, and up to 512GB of unified memory with 1.2TB/s bandwidth. This configuration enables up to 4.3x the peak AI compute performance compared to the M3 Ultra, signaling Apple's strategy to empower developers and researchers to run generative AI workloads directly on their Macs, reducing reliance on hosted APIs.