Apple says the M5 delivers 4X the peak GPU compute for AI compared to the M4. Most tech sites reported the number and moved on. I don’t think people have fully grasped what this means for running AI locally.

The gain isn’t just faster cores. Apple put a Neural Accelerator in every GPU core. I’ve been running local models through MLX on my M2 Mac for a while now, but barely. The M5 turns local AI from “it works, I guess” into something that feels responsive.

There’s a timing angle here too. NAND and memory prices jumped 55-60% in Q1 2026, and the industry expects them to keep climbing. If you want a Mac with serious memory for local AI work, buying now might save you real money over waiting for the M6 or M7. Future machines could carry a much higher price tag for the same RAM.

I look at these numbers, and then I look at my M2 Mac Studio, and I’m raising an eyebrow. The M2 was great when I bought it. But 4X faster prompt processing with purpose-built AI hardware? That’s the kind of gap that makes you start browsing the Apple Store at midnight.

If you have zero interest in local AI, the M5 is just another chip upgrade. But if you’re running models, experimenting with MLX, or thinking about it, this is the first Mac where Apple clearly built the GPU around AI. And with memory prices headed where they’re headed, the window to get in at current pricing might not stay open.