2 M5 Ultra Macs Run Meta's Muse Spark Open Model in Rare Tech Giant Partnership
Meta’s new open-weight AI model is optimized for Apple Silicon, marking a rare technical partnership between the two tech giants for on-device performance.
Key points
- Meta released the Muse Spark Open model, claiming it is the most capable US open-weight model.
- The model runs on two Apple M5 Ultra Mac Studios using the MLX framework.
- CEOs Mark Zuckerberg and John Ternus publicly endorsed the collaboration.
Meta has released a new artificial intelligence model called Muse Spark Open, which the company says can run efficiently on Apple’s latest desktop computers. This release marks a notable shift in the relationship between Meta and Apple, as the two companies have publicly supported each other’s work on this specific technical integration. The model is designed to operate locally on hardware rather than relying solely on remote cloud servers.
The risk
The primary operational consideration here is not a security vulnerability, but a change in how developers deploy large language models. By optimizing Muse Spark Open for Apple Silicon, Meta enables teams to run complex reasoning tasks on local hardware. This reduces dependency on external API providers for certain workloads. However, running such large models requires significant computational resources, specifically the high-end M5 Ultra processors found in Mac Studios.
View the original post on X
View the original post on X
Who is exposed
This development directly impacts software developers and engineering teams who use Apple hardware for local development or inference. According to Meta, the model is particularly suited for complex software engineering and multi-step reasoning tasks. It is also relevant for organizations evaluating on-device AI capabilities to reduce latency or data privacy concerns associated with cloud-based models. The collaboration suggests a growing market for high-performance local AI execution.
Cost and compliance
Meta claims Muse Spark Open is the most capable US open-weight model to date, with 816 billion parameters. The company states it is the first in its class to run on two Apple M5 Ultra Mac Studios using the latest MLX framework. This information comes from Meta’s announcement and Apple’s technical post regarding the array framework for machine learning tasks on Apple Silicon.
What should you do?
Development teams should assess whether their current hardware can support the resource demands of an 816-billion-parameter model. Engineers should review the MLX framework documentation to understand optimization opportunities for Apple Silicon. Organizations should monitor how this partnership influences the broader landscape of open-weight models and on-device AI performance standards.
The background in plain language
Local AI inference means running artificial intelligence models directly on a user’s computer rather than sending data to a remote server. This approach can improve privacy, as sensitive code or data does not leave the device, and reduces latency. However, large models require significant processing power and memory. Frameworks like MLX help translate these complex mathematical operations into instructions that specific hardware, like Apple Silicon, can execute efficiently. Teams usually evaluate trade-offs between cloud convenience and local control when adopting such models.
What to do and how to stay safe: Meta
- Verify that your local hardware meets the minimum specifications for running large-scale models like Muse Spark Open.
- Review the MLX framework documentation to understand how to optimize model performance on Apple Silicon devices.
- Monitor vendor announcements for updates on open-weight model capabilities and compatibility with your existing development environment.
- Assess whether local inference aligns with your organization’s data privacy and security policies for sensitive codebases.
Step-by-step guide: Mobile Security Architecture: 8 Practices for Device Integrity
General security guidance from the Vector Tribune newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
What is the Muse Spark Open model?
It is a new open-weight AI model from Meta, described as the most capable US open-weight model to date, with 816 billion parameters.
How does it run on Mac computers?
It is optimized to run on two Apple M5 Ultra Mac Studios using the latest version of the MLX framework, according to Meta and Apple.
Why is this partnership significant?
Meta and Apple, typically rivals, collaborated over several months to optimize the model for Apple Silicon, with both CEOs publicly endorsing the effort.
