Apple ships its first 2nm chip, M6, and a quad-die M5 Ultra with 512GB unified memory
M6 is Apple's first 2-nanometer chip — a 12-core CPU, 12-core GPU, dual 16-core Neural Engine, and roughly 30% more peak GPU compute for AI than M5 — while M5 Ultra bonds four dies into a single processor with up to 80 GPU cores, 512GB of unified memory, and 1.2TB/s of bandwidth. The memory pool is the real signal: Apple says models with hundreds of billions of parameters run inference entirely on a Mac Studio, putting frontier-scale local inference on a desk instead of a GPU cluster.
Source: apple.com ↗
This lets users store huge datasets entirely in local memory, increase the tokens-per-second speed, and run huge LLMs with hundreds of billions of parameters entirely on device.
Apple
Why this matters
- → 512GB unified memory puts frontier AI models on a desktop instead of a data center
- → 2nm process delivers 30% more GPU compute per watt than previous generation
- → Local inference at scale eliminates cloud dependencies for sensitive workloads
Desktop frontier AI