Meta and Apple Collaborate to Optimize Muse Spark Open Model for Mac Studios
Meta and Apple announce a joint optimization effort for the 800-billion-parameter Muse Spark Open model, enabling desktop execution via the MLX framework on M5 Ultra hardware.
Key points
- Meta released the 816-billion-parameter Muse Spark Open model, claiming it is the most capable US open-weight model available.
- The model runs on desktops using two Apple M5 Ultra Mac Studios with the latest MLX framework, according to Meta.
- Executives from both companies publicly endorsed the collaboration, marking a shift from their typical competitive relationship.
Meta and Apple announced a joint technical optimization effort today, allowing the Muse Spark Open artificial intelligence model to run efficiently on desktop hardware. The companies stated that the model, described by Meta as the most capable US open-weight model to date, can operate using two Apple M5 Ultra Mac Studios. This release marks a significant departure from the firms' usual public rivalry.
The model contains 816 billion parameters and is characterized as a natively multimodal system. Meta claims it excels at long-horizon tasks, complex software engineering, and multi-step reasoning. It is the first model in its class capable of running on desktops using the specified Apple hardware configuration. The optimization relies on the latest version of Apple's MLX framework, an array framework designed for machine learning tasks on Apple Silicon.
View the original post on X
View the original post on X
Executive Endorsement and Collaboration
Meta CEO Mark Zuckerberg posted on X that the model runs great on new Mac Studios thanks to work with Apple CEO John Ternus. Ternus replied to congratulate the team, noting the efficiency of running a model of this caliber directly from a desk. He stated that the Mac Studio continues to push the boundaries of on-device artificial intelligence. These public exchanges highlight a coordinated marketing and technical launch strategy.
According to Axios, Meta chief AI officer Alexandr Wang commented on the improved relationship between the two tech giants. Wang stated, “We’re excited for mommy and daddy to be playing nice again.” Axios reported that the companies have been collaborating over several months to achieve this technical integration. This follows earlier reporting by Threat Times on October 15, 2026, which noted Wang’s claim that former Apple executive Daniel Gross helped bridge the gap between the firms.
Technical Requirements and Availability
Apple published a new technical post detailing the MLX framework's performance with Muse Spark Open and two other open models. The setup requires two Apple M5 Ultra Mac Studios to run the model locally. Earlier Threat Times reporting indicated that the model requires a terabyte of RAM to run locally, a specification consistent with the high-end hardware requirements. The weights for the model were released on the same day as the announcement.
What should you do?
Security and technology professionals should monitor the performance and resource consumption of large open-weight models running on local hardware. Organizations evaluating on-device AI should verify the specific hardware requirements, such as the need for multiple high-end Mac Studios, before deployment. IT teams should assess the security implications of running large models locally, including data privacy and potential side-channel risks.
What to do and how to stay safe: Meta
- Verify that local hardware meets the specific memory and processing requirements for large open-weight models before attempting deployment.
- Review the security implications of running multimodal AI models on desktop systems, particularly regarding data handling and isolation.
- Monitor vendor updates for the MLX framework and associated models to ensure compatibility with current security patches.
- Assess the operational costs of maintaining the necessary high-end hardware infrastructure for local AI inference tasks.
Step-by-step guide: Mobile Device Security: How to Protect Data Beyond Passwords
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
What hardware is required to run the Muse Spark Open model?
The model requires two Apple M5 Ultra Mac Studios running the latest MLX framework to operate on desktop hardware.
How large is the Muse Spark Open model?
The model has 816 billion total parameters, making it Meta's largest US model to date.
What is the MLX framework?
MLX is an array framework for machine learning tasks on Apple Silicon, optimized for running models like Muse Spark Open.
