Meta's New AI Runs on Your Mac, But Needs a Huge Machine

Meta has released a new AI model called Muse Spark Open. It's an open-weight model, meaning the underlying code is publicly available for anyone to use or modify. The big selling point: it's designed to run directly on Apple Macs, not just in massive data centers. This could let you run advanced AI on your own computer, keeping your data private and saving on cloud costs.

However, there's a catch: to run it locally, you need a Mac with at least 1 terabyte of RAM. That's an enormous amount of memory—most laptops have 8-16GB. Only Apple's highest-end machines, like the Mac Studio with the M5 Ultra chip, can handle it. So while it's technically possible to run on a Mac, it's not for the average user. It's aimed at professionals or companies with powerful hardware.

What can it do? Meta says it excels at coding and agentic tasks—AI that can take actions on your computer, like fixing bugs in apps or editing video. They tested it on a Mastodon client, where it made bug fixes and added features, and on DaVinci Resolve, where it reviewed and cropped footage. This shows it can handle real-world, complex jobs that usually require human effort.

Open-weight models like this offer advantages: they can be cheaper to run, more transparent, and customizable. Meta is making it available through various API platforms, so developers can integrate it into their own apps. For everyday people, this means more powerful AI tools might become available, but the hardware requirements mean it won't be on your typical laptop anytime soon. It's a step toward decentralized, private AI, but the barrier to entry is still high.