Apple and Meta have announced a collaboration aimed at enabling developers to run Meta’s sophisticated AI model, Muse Spark Open, directly on high-end Mac computers. This development signals a growing industry trend toward executing powerful artificial intelligence models using local hardware rather than depending solely on cloud data centers.
The Nature of the Collaboration
The partnership involves Meta’s new open-weight AI model, Muse Spark Open. According to Meta, this model is the most advanced open-weight system available under a license that permits limited commercial use. Technically, it is an 81.6-billion-parameter model built using a mixture-of-experts architecture. As an open-weight resource, developers are able to download the model’s parameters, or weights, from the internet to operate it on a variety of hardware setups.
Using Muse Spark, customers will be able to pair between two and four Mac Studio PCs to perform tasks such as coding, video analysis, or other agentic work. The companies stated that running the model locally, rather than in the cloud, provides both enhanced security and cost savings for users. This announcement, following several months of joint effort, suggests that the AI boom has begun to reshape historical alliances within the technology sector.
Rivalries and Industry Shifts
The alliance represents an unlikely pairing, given the long-standing professional rivalry between Apple and Meta. Historically, the two companies often clashed, particularly when Tim Cook was leading Apple. The current collaboration is credited with finding common ground between the two tech giants, with John Ternus currently at the helm of Apple. Representatives from both companies declined to confirm whether the Muse Spark Open partnership marks the beginning of a deeper, sustained relationship.
The history of friction is notable; the relationship significantly deteriorated after Apple implemented privacy changes in 2021 that limited the ability of digital advertising companies to target users. Meta estimated in early 2022 that this change alone would reduce its revenue by $10 billion.
Alexandr Wang, Meta AI chief, noted that many powerful open-weight models are too large to run on most personal computers, adding that Apple’s Mac Studios provide sufficient memory to make running Muse Spark Open practical. This move is viewed as a significant display of confidence in Meta’s renewed AI efforts and could potentially challenge established leaders like OpenAI and Anthropic. It also underscores Apple’s increasing strength in AI chips, a market previously dominated by Nvidia in large data centers.
Technical Requirements and Availability
Developers can access the AI model by downloading it from platforms like Hugging Face without needing specialized configurations for Apple’s silicon. Muse Spark Open has also been optimized for faster performance on Apple hardware. Apple subsequently published new benchmarks for its chips and technical documentation regarding its MLX AI framework, detailing its ability to distribute open-source models across multiple machines.
However, the hardware required for this local setup is substantial. The model demands Apple’s fastest chips and the maximum amount of memory. Currently, a Mac Studio equipped with the necessary M5 Ultra chip and 256GB of memory costs at least $10,799. A higher-end version featuring 512GB of RAM is scheduled for release later this month, and users must account for purchasing multiple machines.
Broader Context in AI Development
The collaboration focuses on open-source AI, a topic of increasing scrutiny in Washington, where politicians are concerned about foreign models potentially undermining American AI development. While Meta previously anchored its AI strategy on its older open-weight Llama family of models, it has since shifted its focus toward creating proprietary models under Wang’s leadership. Wang joined Meta after the company invested over $14 billion in his previous startup, Scale AI, last year.
While Apple and Meta (formerly Facebook) had previously partnered on iPhone extensions more than a decade ago, they have had no public partnerships since 2017, following Apple’s removal of system-level Facebook integration from the iPhone. In discussing the collaboration, Wang mentioned that while “Most of the animosity predates my tenure at the company, but obviously there’s been a long history between Meta and Apple,” pointing to Daniel Gross, an AI researcher and former Apple director, as an example of the renewed professional ties.
Meta is not the first AI developer to partner with Apple. In 2024, Apple announced a partnership with OpenAI to integrate ChatGPT into Siri, though OpenAI noted in a legal filing regarding a dispute with Elon Musk’s xAI that the integration “dramatically underperformed.” Furthermore, Apple recently announced a deal with Google to incorporate its Gemini models into Apple Intelligence, alongside using Google Cloud infrastructure.
Despite these moves, Apple’s AI strategy has faced criticism from investors and customers, partly because the updated version of Siri, released recently, is not as advanced as leading chatbot and agent systems. Meanwhile, Meta’s business model remains heavily dependent on advertising revenue, leading to continued scrutiny. Recent developments, including Meta’s Muse personal agent app and its proprietary foundation model, Muse Wonder, have generated optimism among investors regarding new AI business opportunities.
Regarding Muse Spark Open, Wang stated, “These aren’t the most capable models that we’ve developed or released… But we think that there will be a lot of companies and developers who sort of use this model and tinker with it at small scale and then seek to ramp up and scale up their usage.”
