Surprise collaboration from Meta and Apple: Giant AI model will run on Mac computers
Meta has introduced its 816-billion-parameter AI model, Muse Spark Open. It can run locally on Mac Studio via Apple Silicon integration.

Meta has introduced Muse Spark Open, its new artificial intelligence model with 816 billion parameters. Thanks to a technical collaboration with Apple, the model can run locally on two Mac Studio computers equipped with M5 Ultra processors.
Meta, one of the leading companies in the AI industry, announced the next-generation Muse Spark Open model on October 15, 2026.
Released by the company with open weights, the model is specifically designed for software development, multi-step tasks, and solving complex problems.
816 Billion Parameters and Advanced Reasoning
Muse Spark Open has a total of 816 billion parameters. Thanks to its mixture-of-experts architecture, approximately 42 billion parameters are activated during processing.
The model features a one-million-token context window. This capacity aims to facilitate the analysis of long documents and large software projects.
According to Meta's statement, the model features a multimodal architecture capable of processing different data types, primarily text.
Collaboration with Apple
Meta utilized Apple's MLX framework to enable the Muse Spark Open model to run on Apple Silicon processors.
The new AI model can be run locally using two M5 Ultra Mac Studios, each with 512 GB of unified memory.
Thus, it becomes possible to run a high-capacity AI model on powerful desktop hardware outside of dedicated cloud servers.
New Opportunities for Developers
Meta states that the model can be used in software development and autonomous task execution scenarios.
Code editing, debugging, long-term software operations, and various automation processes are among the targeted use cases for the model.
Making Muse Spark Open available to developers with open weights provides new opportunities for customizing AI solutions.
However, since running this model locally at this scale requires high memory capacity and powerful hardware, it is not an option accessible to every user.
The collaboration between Apple and Meta is seen as a striking step in the competition to run powerful artificial intelligence systems on personal computers.



