Apple and Meta Join Forces to Run Muse on Macs

Deep News10-16 08:16

Apple and Meta have announced a partnership to bring Meta's most powerful open-source AI model onto Apple's high-end desktop computers, marking a moment when two longtime Silicon Valley rivals set aside historical grievances and aligned themselves on the same side amid the artificial intelligence wave. This collaboration not only serves as a public endorsement of Apple's AI chip capabilities but also poses a potential challenge to OpenAI and Anthropic.

According to CNBC, the two companies announced on Thursday (October 15) that developers will be able to run Meta's latest open-source model Muse Spark Open locally on Apple Mac Studio computers. Users can combine two to four Mac Studios to complete coding, video analysis, and other agentic tasks, with the model running locally rather than relying on the cloud, balancing security and cost control. Meta AI chief Alexandr Wang said that Apple's Mac Studio offers sufficiently large memory to make running Muse Spark Open on a personal computer a reality.

The partnership comes just one month after former Apple CEO Tim Cook stepped down and John Ternus took over the helm at Apple. Ternus responded to Zuckerberg's post on social media, saying, "It's amazing to be able to run a model of Muse Spark Open's caliber on the desktop so efficiently. Mac Studio keeps pushing the boundaries of on-device AI." This interaction is seen by observers as a clear signal of warming relations between the two companies.

However, the hardware requirements for this solution are quite steep. Currently, a Mac Studio equipped with the M5 Ultra chip and 256GB of memory starts at $10,799, and a higher-spec version with 512GB of memory will be released later this month. In addition, fully running Muse Spark Open requires purchasing multiple machines at the same time.

Two Rivals Shake Hands: AI Reshapes the Silicon Valley Landscape

The rift between Apple and Meta dates back a long way. In 2021, Apple introduced a new privacy policy that limited digital advertising companies' targeting capabilities, and Meta (then known as Facebook) promptly estimated that the change would reduce its 2022 revenue by about $10 billion, pushing relations to a freezing point.

Previously, Cook had publicly mocked Zuckerberg during the 2018 Cambridge Analytica scandal, saying, "I wouldn't be in this situation"; Zuckerberg said in an interview with Joe Rogan last year that Apple "hasn't invented anything great since the iPhone." The last public collaboration between the two companies was in 2017, when Apple removed Facebook system-level integration from the iPhone.

After Ternus took office, the situation began to loosen. In an interview, Wang partly credited the partnership to his team member Daniel Gross, a former Apple AI research director—both Gross and former GitHub CEO Nat Friedman were important talents Zuckerberg aggressively recruited last year, when Zuckerberg was fully restructuring Meta's AI strategy. Wang admitted, "Most of the bad blood happened before I joined, but there is indeed a long history between Meta and Apple."

Muse Spark Open: An Open-Source Flagship with 81.6 Billion Parameters

Muse Spark Open is Meta's most powerful open-source model to date, using a mixture-of-experts architecture with 816 billion parameters, released under a license that allows limited commercial use. As an open-weight model, developers can download the model weights directly from platforms such as Hugging Face and run them on Apple chips without additional configuration.

Meta also specifically optimized and adjusted the model for Apple hardware, making it run faster on Apple devices. Apple simultaneously released new chip benchmark data and made public technical documentation on how its MLX AI framework splits open-source models across multiple machines.

Wang said, "We worked with Apple to bring leading agentic capabilities to the desktop for the first time, and the Mac Studio with Apple's M5 Ultra chip is key to achieving that." He also noted that Muse Spark Open is not Meta's most advanced model, but "a large number of companies and developers will use this model for small-scale exploration and then seek to scale up their usage."

Apple's AI Landscape: Multiple Partnerships, Pressure Remains

Meta is the latest piece in Apple's recent strategy of broad AI alliances. In 2024, Apple announced a partnership with OpenAI to integrate ChatGPT into Siri, but according to documents disclosed by OpenAI in legal disputes related to Musk's xAI, the integration "performed significantly below expectations" and failed to drive OpenAI user growth. Apple subsequently sued OpenAI in July this year for trade secret theft, and the latter denied the allegations. Earlier this year, Apple also announced an agreement with Google to incorporate the Gemini model and Google Cloud infrastructure into the Apple Intelligence system.

Despite an expanding roster of partners, Apple's AI strategy still faces criticism from investors and users—the new version of Siri released last month is widely considered to still lag behind the chatbots and agents of top AI labs.

For Meta, the partnership is an important endorsement of its AI strategy revival. Meta earlier this week released its latest proprietary foundation model Muse Wonder, after previously launching the personal agent app Muse and the small open-source model Muse Glimmer unveiled in August. This series of moves has begun to boost investor sentiment, with market expectations gradually rising for Meta to open up new AI business opportunities. However, Meta's revenue sources are still almost entirely dependent on its advertising business.

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Comments

We need your insight to fill this gap
Leave a comment