Apple has introduced two new Mac chips, the M6 and the M5 Ultra, with both aimed squarely at running AI work on the machine in front of you rather than in a distant data center.
According to News9Live, the M6 is Apple's first chip built on a 2nm design — a reference to how small the transistors are, with smaller generally meaning more performance for the same power. The same report says the pair was introduced with a stronger focus on local AI and "agentic" workloads, meaning AI that carries out multi-step tasks on its own.
The M5 Ultra is the headline part for anyone with big models to run. ZDNet reports it is Apple's most powerful chip ever, with AI performance four times faster than the M3 Ultra it succeeds. In the Mac Studio, ZDNet says the M5 Ultra will offer up to 512GB of unified memory and can drive up to eight external displays. Engadget puts an 80-core GPU in the chip, framing it as built to power through AI models and 8K video. Digital Trends describes the launch simply as more horsepower and more AI compute.
The memory figure is the part worth dwelling on. Unified memory is shared between the processor and graphics side, so on Apple's design it doubles as the pool an AI model has to live in. A 512GB ceiling means models that would otherwise require server hardware can, in principle, be loaded on a desktop.
Engadget has published a direct M5 Ultra versus M3 Ultra comparison for readers weighing an upgrade.
It matters because the center of gravity for AI is shifting: if the largest models can run locally, your data — and your bill — may not have to leave the room.