Apple and Meta team up to run new Muse Spark Open AI model on Mac Studio

Long-time rivals collaborate to bring an 816-billion-parameter open-weight model to the desktop

By LineZotpaper
Published
Read Time2 min
Sources3 outlets
Apple and Meta announced a collaboration Thursday that will let developers run Meta's new open-weight AI model, Muse Spark Open, locally on high-end Mac Studio computers. The partnership, reached after months of work, marks a notable thaw in relations between two companies that have clashed for years over privacy and advertising.

The collaboration centers on Muse Spark Open, which Meta describes as the most capable open-weight model released under a license allowing limited commercial use. It is an 816-billion-parameter model built on a mixture-of-experts architecture, and as an open-weight model its parameters can be downloaded and run across a wide range of hardware.

Customers will use the model for tasks such as coding, video analysis and agentic work. The companies said running the model locally instead of in the cloud adds a layer of security and saves users money.

The Verge reports that Meta calls Muse Spark Open the "most capable US open-weight model to date" and "the first model in its class" to run on a desktop, using two Apple M5 Ultra Mac Studios with the latest MLX framework. CNBC reports that customers can pair between two and four Mac Studio computers for the task.

The announcement is the latest sign that the AI boom is shifting historical alliances in Silicon Valley. Apple and Meta have long regarded each other warily, with relations reaching a low point after Apple introduced privacy changes in 2021 that limited ad targeting. Meta estimated in early 2022 that the change would cut its revenue that year by $10 billion.

Meta CEO Mark Zuckerberg posted on X: "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."

Representatives from both companies declined to say whether the collaboration signals a deeper relationship.

§

Analysis

Why This Matters

  • Two of Silicon Valley's most prominent rivals are publicly collaborating, a sign that competitive pressure in AI is redrawing industry alliances.
  • The move pushes large-model computing toward the desktop, with implications for cost, privacy and how developers deploy AI workloads.
  • It strengthens the open-weight model movement at a time when much of the industry is focused on closed, cloud-based AI offerings.

Background

Apple and Meta have competed for years in smartphones, advertising and virtual reality. Their relationship soured sharply after Apple's 2021 app tracking transparency changes, which Meta estimated would cost it billions in ad revenue. Apple's shift to its own silicon has made its high-end Macs increasingly capable for AI work, and Meta has pursued an open-weight strategy aimed at winning developers. This collaboration brings those two trajectories together under new leadership at Apple.

Key Perspectives

Apple: The collaboration showcases the Mac Studio and M5 Ultra as a platform for serious AI work, and new CEO John Ternus is publicly embracing the partnership. Meta: The open-weight release gives developers a commercially usable, downloadable alternative to closed models, and Zuckerberg is personally promoting it. Critics/Skeptics: The license only allows limited commercial use, raising questions about how "open" the model really is. The practical cost of pairing multiple Mac Studios may also limit adoption, and the two companies' history suggests the relationship could remain transactional.

What to Watch

  • Developer uptake and independent benchmarks comparing Muse Spark Open with other leading open-weight models.
  • Whether Apple and Meta announce further joint projects or this remains a one-off collaboration.
  • The terms of the license and how "limited commercial use" is enforced in practice.

Sources

Zotpaper

Written by software from the reporting listed above, scored by an automated standards desk, and published without a person reading it first. If something here is wrong, tell the editor and it will be put right.

How we workSubscribe