Owning Your AI No Longer Takes a Data Center
Owning your AI used to mean owning a data center. Meta's new open model runs on two desktop Macs, though the price, the license and Meta's control still matter.
Owning your own AI used to mean owning a data center. Meta just shrank that to two desktop computers. On October 15, the company released Muse Spark Open, a large AI model that anyone can download, and showed it running on two linked Apple Mac Studios.
Meta calls it the most capable open model built by a U.S. company. It is designed for coding and for agents that carry out long, multistep jobs, and it can read text, images, audio and video. In Apple's own tests, two Macs produced about 16 to 25 tokens, or word fragments, per second, and four Macs about 37.
Data That Stays Home
The appeal is control. When a hospital, bank or ministry runs the model on machines it owns, its documents and questions never have to leave the building. No outside company sees them, and no one else can switch the service off.
The Fine Print
It is not cheap. Apple has not yet priced the 512GB Mac Studio the setup needs, which arrives in late October, and VentureBeat estimates comparable hardware at tens of thousands of dollars. Meta's own scores also show it trailing several Chinese open models on coding tests.
"Open" has limits, too. The license is free for most uses, but companies with more than $20 million in revenue that sell access to the model as a service need a separate deal with Meta. The usage rules also ban military and espionage work and running critical infrastructure, which matters for governments. On Hacker News, where the launch drew more than 900 points, many commenters debated whether it counts as open at all. And Meta still decides how the model is trained and whether a next version ever ships.
For companies and governments, the move is simple. Test it on a real, sensitive task, have a lawyer read the license, and keep a backup model ready. Owning the hardware is now the easy part. Owning the model's future is not.
Sources
- Meta AI Research, Introducing Muse Spark Open, The Most Capable US Open-Weight Model to Date, October 15, 2026
- Apple Machine Learning Research, Local and Distributed Large LLM Inference with MLX and RDMA on Mac Studio, October 15, 2026
- Hugging Face, Muse Spark Open model card and Muse Model License, October 15, 2026
- The Verge, Meta and Apple team up to make Muse AI models run better on Macs, October 15, 2026
- VentureBeat, Meta unveils Muse Spark Open, the new top U.S. open weights model, and it runs on two paired Mac desktops, October 15, 2026
- Apple Newsroom, Apple introduces new Mac Studio with M5 Max and M5 Ultra, August 25, 2026
- Hacker News, Muse Spark Open discussion thread, October 2026