According to Zhitong Finance APP, Meta (META.US) and Apple (AAPL.US) announced a new partnership on Thursday, enabling developers to easily run Meta’s newly released open-weight AI models on Apple’s high-end desktop computers.
According to Meta, the core of this partnership is Muse Spark.
Open is the most powerful open-weight model released by the company under a license that permits limited commercial use. It is an 816-billion-parameter model built on a so-called mixture-of-experts architecture. As an open-weight model, developers can download its parameters (i.e., weights) from the internet and run the model on various hardware platforms.
The two companies said users can network two to four Mac Studio units using Muse.
Spark performs programming, video analysis, or other intelligent agent tasks. The model runs locally on your device without requiring cloud upload, enhancing data security and reducing cloud service costs for users.
The partnership, officially announced on Thursday after months of refinement, marks how the AI boom is reshaping the long-standing rivalries in Silicon Valley. For years, Apple and Meta have been wary of each other and had virtually no public collaboration. When Apple introduced privacy updates in 2021 that significantly weakened digital advertisers’ targeting capabilities, their relationship hit a low point. In early 2022, Meta estimated that this change would reduce its revenue by $10 billion that year.
Representatives from Apple and Meta declined to disclose whether the Muse Spark Open collaboration signals a deeper strategic partnership between the two companies.
Zuckerberg posted a screenshot on X showing Meta's AI agent running on a Mac, writing: “Muse Spark”
Open—Weight downloads are officially available today. Thanks to the collaboration with @johnternus and the Apple team, this model performs exceptionally well on the new Mac Studio.
Tenus responded: "Congratulations to the team! It's incredible to see a model of Muse Spark Open's scale and complexity running so efficiently on desktop. Mac"
Studio continues to push the performance boundaries of on-device AI!
Alexander Wang, head of Meta AI, said in an interview that many powerful open-weight models are too large to run on most personal computers, but Apple’s Mac
Studios provides users with sufficient memory to make the local deployment of Muse Spark Open feasible.
This partnership demonstrates strong confidence in Meta’s renewed focus on AI and presents a potential competitive challenge to OpenAI and Anthropic. It also signals that Apple’s AI chip capabilities continue to strengthen. Currently, NVIDIA (NVDA.US) dominates the AI chip market through its widespread deployment in large data centers.
We partnered with Apple to bring the first Mac equipped with the M5 Ultra chip
“We are thrilled to bring cutting-edge intelligent agent capabilities to Studio desktop devices,” Wang said, noting that developers can now access Hugging
Download the model on platforms like Face, and it will run natively on Apple’s custom chips without additional adaptation. Muse Spark
Open has been optimized and accelerated for Apple hardware. Apple has also released new benchmark data for its chip and published technical documentation detailing how the MLX AI framework splits open-source models and runs them distributively across multiple machines.
However, the Mac Studio is expensive, and this AI model has strict requirements for Apple's top-tier chip and large amounts of memory.
Currently, the Mac equipped with the required M5 Ultra chip and 256GB of memory
The Studio starts at a minimum price of $10,799. The high-end model with 512GB of storage will be available later this month.
In addition, users need to purchase multiple server units to set up a network.
Reach consensus in the open-source AI field
Two companies are collaborating in the open-source AI space, a field that has become a point of contention in Washington as policymakers fear that China’s free open-source models could undermine U.S. AI. Although Meta’s early AI strategy centered on its Llama series of open-weight models, the company has since shifted its focus under Alexander Wang’s leadership toward developing proprietary models. Alexander Wang joined Meta last year, when the company was transitioning from its previous startup, Scale.
AI has invested over $14 billion.
Although Apple and Meta (then known as Facebook) collaborated on iPhone features over a decade ago, no public partnership has been reached since Apple removed system-level Facebook integration from the iPhone in 2017.
Most of the conflicts between the two parties existed before my arrival. The tensions between Meta and Apple have been longstanding,” Wang said, noting that one member of his team—AI researcher and former Apple director Daniel Gross—was one of the factors that helped facilitate this partnership. Gross, a former GitHub
Like CEO Nat Friedman, he was a key hire when Zuckerberg overhauled the company’s AI strategy last year.
There have also been sharp verbal exchanges between the leaders of both sides.
During the 2018 Cambridge Analytica scandal, Cook, then Apple’s CEO, subtly mocked Zuckerberg, saying, “I wouldn’t find myself in that situation.” Cook has frequently criticized Meta for undermining user privacy. (Tenus succeeded Cook as Apple’s CEO last month.)
Zuckerberg has also publicly criticized Apple, stating on a podcast last year that Apple "hasn't invented anything great since the iPhone." He has long sought to build a platform fully controlled by Meta to reduce dependence on Apple.
Tenus, a hardware engineer by background, has already begun adjusting Apple’s product release schedule. Meanwhile, Meta’s internal developers also favor Apple devices. Earlier this year, Meta purchased hundreds of Macs for its Superintelligence team to conduct local AI and intelligent agent experiments.
Apple's multi-party AI partners
Meta is the latest AI model developer to partner with Apple.
In 2024, Apple announced a partnership with OpenAI to integrate ChatGPT into Siri; however, OpenAI stated that the integration "performed far below expectations" and did not drive user growth for OpenAI. This claim comes from a filing submitted by OpenAI in its legal dispute with Elon Musk’s xAI. Apple sued OpenAI in July, accusing it of stealing trade secrets, a claim OpenAI has denied.
Earlier this year, Apple announced an agreement with Google (GOOGL.US) to integrate the Gemini model into Apple Intelligence, using Google Cloud infrastructure.
Nevertheless, Apple's AI strategy has faced criticism from investors and customers, partly due to its newly released version of Siri last month, which lacks the capabilities of leading laboratory chatbots and intelligent agents.
Meanwhile, Meta is fully pushing into the top AI lab赛道, but has consistently struggled to keep pace with competitors, and its business still relies almost entirely on advertising. Recent developments include Meta’s Muse personal AI agent app and its latest closed-source foundation model, Muse, released earlier this week.
Wonder has excited investors who are optimistic about Meta AI's new business opportunities.
Open source remains a key component of Zuckerberg's AI strategy. In August, Meta launched a model called Muse
Glimmer's small open-weight AI model; Zuckerberg has pledged to eventually release larger, more powerful products.
Muse Spark Open's performance is inferior to the proprietary product Muse Wonder, but Wang stated that it is sufficient to attract more developers to upgrade their hardware and give it a try.
“These are not the most powerful models we have developed or released,” Wang said, “but we believe many companies and developers will use and fine-tune this model on a small scale before expanding deployment.”