Apple and Meta partner to run massive Muse AI model locally on Mac Studios

Apple unveils the new Mac Studio, the most powerful Mac ever, featuring M5 Max and M5 Ultra.
Apple’s new Mac Studio, the most powerful Mac ever, featuring M5 Max and M5 Ultra.

Mac Studio delivers extensive pro connectivity, including up to six ports of Thunderbolt 5, support for up to eight displays, and now Wi-Fi 7 and Bluetooth 6.
Mac Studio delivers extensive pro connectivity, including up to six ports of Thunderbolt 5, support for up to eight displays, and now Wi-Fi 7 and Bluetooth 6.

Apple and Meta announced a collaboration on Thursday that will let developers run Meta’s large Muse Spark Open model locally on high-end Mac computers, an unusual alliance between the two longtime rivals.

The partnership centers on Muse Spark Open, an 816-billion-parameter open-weight model that uses a mixture-of-experts architecture, CNBC reports. Developers can download the model’s weights and run it on pairs of two to four Mac Studio machines equipped with Apple’s M5 Ultra chips and high amounts of unified memory. The setup is designed for coding, video analysis, and other agentic tasks, with Apple optimizing performance through its MLX framework for distributed inference across machines.

Meta CEO Mark Zuckerberg wrote on X that “Muse Spark Open — weights releasing today — runs great on new Mac Studios thanks to work with @johnternus and Apple.” Apple CEO John Ternus replied, “Congrats to the team! Amazing to run a model of Muse Spark Open’s caliber and scale this efficiently, right from your desk. Mac Studio keeps pushing the boundaries of on-device AI!”

Meta’s AI chief Alexandr Wang said the companies had been working together for months, with former Apple executive Daniel Gross, now at Meta, playing a key role in the collaboration. “We worked with Apple to bring leading Agent capabilities to your desktop for the first time on the Apple M5 Ultra Mac Studios,” Wang said. He added that while the model is not Meta’s most capable, it is expected to attract developers who want to experiment at smaller scale before expanding.

The companies have a history of friction. They had not publicly partnered since around 2017, and tensions included Apple’s 2021 privacy changes that hurt Meta’s advertising business as well as public criticism from former Apple CEO Tim Cook and Zuckerberg. The current effort coincides with Ternus’s leadership at Apple. Representatives from both companies declined to say whether the work signals a broader ongoing relationship.

Running the model locally is intended to offer advantages in security, cost, and reduced reliance on cloud data centers. A Mac Studio with M5 Ultra and at least 256 GB of memory starts at $10,799, and multi-machine configurations raise the expense further. Meta has also purchased hundreds of Macs earlier this year for its own AI experimentation.

The announcement fits into Apple’s wider AI approach, which includes partnerships with OpenAI and Google for features in Apple Intelligence and Siri, while Meta continues advancing both proprietary and open-weight models under Wang.

It also highlights growing interest in powerful on-device AI as an alternative to cloud-only systems.
‎
MacDailyNews Take: As Lord Palmerston said, “There are no permanent enemies, and no permanent friends, only permanent interests.”

More info about Muse Spark Open: research.meta.ai/blog/muse-spark-open


‎
Please help support MacDailyNews — and enjoy subscriber-only articles, comments, chat, and more — by subscribing to our Substack: macdailynews.substack.com. Thank you!

Support MacDailyNews at no extra cost to you by using this link to shop at Amazon.

1
2

2 Comments

  1. you need to (2) Mac studios with 5128 GB each to run the full model. its a 850 GB model.

    “Apple’s unified memory architecture lets the M5 Ultra’s integrated GPU use up to 512 GB per Mac Studio, so two machines can hold an entire model. “

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.