Meta and Apple have collaborated to enable Meta's newly introduced Muse Spark Open artificial intelligence model to run on Apple's Mac Studio computers, marking a technical partnership between two companies that have previously disagreed over privacy and platform policies.
Meta's AI chief Alexandr Wang disclosed that Daniel Gross, a former Apple executive who now works at Meta, played a significant role in facilitating the collaboration. Speaking to Axios, Wang said the two companies had been working together for several months to support the deployment of Meta's AI model on Apple's hardware.
The collaboration enables Muse Spark Open, an 816-billion-parameter model, to operate across two or four interconnected Mac Studio computers powered by Apple's M5 Ultra chips. The system uses Apple's MLX framework to distribute the model's computational requirements across multiple machines.
Muse Spark Open is built using a mixture-of-experts architecture and is designed to support tasks including complex reasoning, extended coding sessions and video analysis. Meta has described it as its most capable open-weight AI model, developed in the United States.
The model is available for download through platforms including Hugging Face, with licensing provisions permitting commercial applications subject to certain restrictions.
According to Wang, Gross helped establish cooperation between the companies despite their history of disagreements.
“I think he played a really big role in kind of getting the two companies to work nicely together,” Wang told Axios.
Apple has also released technical documentation and performance benchmarks explaining how its MLX framework manages large AI models across connected Mac Studio systems.
One aspect of the deployment is that developers can operate the model locally rather than sending their prompts and files to Meta's servers. This arrangement allows organisations and individual users to retain greater control over the information processed through the model.
“Meta has no control or visibility to any of that,” Wang said, referring to the information processed through locally operated systems.
However, running the model requires substantial hardware investment. According to the reported specifications, an M5 Ultra Mac Studio configured with 256GB of memory costs more than ₹9 lakh when converted from US pricing, and the deployment requires at least two machines.
Apple is also expected to introduce a 512GB configuration later in October, providing additional memory capacity for demanding AI workloads.
The collaboration follows years of public disagreements between Apple and Meta, particularly over data privacy and digital advertising.
Their differences became prominent during the Cambridge Analytica controversy in 2018, when Apple leadership criticised Facebook's approach to user information.
Tensions continued in 2021 after Apple introduced privacy controls that limited how applications could track users across third-party services, affecting advertising measurement and targeting practices.
The latest collaboration centres on technical compatibility rather than a broader commercial partnership. Neither company has announced additional joint products or services arising from the arrangement.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.