Meta and Apple partner to run Muse Spark Open AI model locally on Mac computers

Meta and Apple have collaborated to enable Meta's new Muse Spark Open AI model to run locally on high-end Mac Studio systems using Apple silicon.

Meta has released Muse Spark Open, a natively multimodal open-weight model designed for complex coding and agentic tasks. According to Meta, the model features 816 billion parameters and a 1 million token context window. TechCrunch reported that the model is designed to run locally on Apple hardware, though it requires significant power, including at least 1 terabyte of RAM.

The collaboration allows developers to run the model on Mac Studio desktops, specifically those equipped with the M5 Ultra chip. Users can reportedly pair two to four Mac Studio PCs to handle tasks like video analysis and coding. Meta demonstrated the model's capabilities by using it with Muse Code to fix bugs in a Mastodon client and using it within DaVinci Resolve to review and crop footage.

The International reported that this partnership follows months of collaboration between the two tech rivals. Mark Zuckerberg stated on X that the weights for the model are being released today, noting the system runs efficiently on Mac Studios thanks to work with Apple. John Ternus of Apple responded that the Mac Studio is pushing boundaries by running a model of this caliber right from a user's desk.

While the model is available through various API platforms, running it locally offers advantages in cost-efficiency and security compared to cloud-based systems. Meta AI chief Alexandr Wang told the International that while many powerful models are typically too large for personal computers, Apple’s hardware provides enough memory to make running Muse Spark Open practical.

Background

The partnership marks a shift in the relationship between Meta and Apple, which was previously strained by 2021 privacy changes that Meta estimated would cost it $10 billion in revenue.