Meta and Apple Join Forces: Muse Spark Open Runs Locally on Mac Studio

Meta today released the Muse Spark Open AI model, billing it as the most capable open-weight model from the US to date, and worked with Apple to make it run locally on Mac Studios powered by Apple's M5 Ultra chips through the MLX framework. The two companies' CEOs publicly backed each other, marking a rare collaboration between longtime rivals in AI. ^[1]

The launch: 816 billion parameters, natively multimodal

Muse Spark Open, as Meta announced it, is a natively multimodal model with 816 billion total parameters. According to Meta, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning. It is what the company calls 'the most capable US open-weight model to date.' ^[1] That parameter count puts it at the top tier of open models, meaning it can handle tasks that demand large amounts of context and complex logic, while open weights let developers and researchers download and fine-tune it freely, potentially spurring more localized AI applications.

The Apple partnership: from rivals to collaborators

Meta and Apple, typically rivals, worked together on this launch and are publicly supporting each other. Meta CEO Mark Zuckerberg said on X that Muse Spark Open's weights ship today and run great on new Mac Studios 'thanks to work with @johnternus and Apple.' New Apple CEO John Ternus replied to congratulate the team, saying it is 'amazing to run a model of Muse Spark Open's caliber and scale this efficiently, right from your desk' and that 'Mac Studio keeps pushing the boundaries of on-device AI.' ^[1] Meta chief AI officer Alexandr Wang told Axios that the two companies have worked together for months. The cooperation breaks with the two firms' usual competitive posture in consumer electronics and AI.

Technical details: the MLX framework and local inference

The model is the first in its class that can 'run on your desktop using two Apple M5 Ultra Mac Studios with the latest MLX framework.' Apple also published a new technical article about MLX — the company's array framework for machine learning on Apple Silicon — covering Muse Spark Open and two other open models. ^[1] That means developers can run large-scale AI models efficiently on local hardware rather than depending on the cloud, potentially cutting inference costs and improving privacy.

Industry impact: open weights meet on-device AI

Meta calls Muse Spark Open 'the most capable US open-weight model to date,' a positioning that underscores intensifying competition among open AI models. At the same time, the Apple partnership shows that on-device AI is becoming a major path to deployment for large models. With Mac Studio's high-performance chips and the MLX framework, an 816-billion-parameter model can run on a desktop — something that could encourage more companies to explore local AI inference and push hardware-model co-optimization. ^[1]

What to watch next

Whether the Meta-Apple collaboration lasts, and how Muse Spark Open performs in real-world use, will be an important window into how open AI and on-device computing converge. With the weights now out, feedback and fine-tuning results from the developer community are also worth watching. ^[1]