Meta and Apple collaborate to enable local Muse Spark Open on M5 Ultra Mac Studios
The tech giants partnered to run Meta’s large open-weight model locally on Apple Silicon, allowing users to link multiple machines for complex agentic tasks.
Key points
- Meta and Apple engineered Muse Spark Open to run locally on M5 Ultra Mac Studios.
- Users can pair two to four Mac Studios to handle coding and agentic workloads.
- The collaboration marks a shift in AI strategy between the historically rival companies.
Meta announced that it has worked directly with Apple to enable its Muse Spark Open model to run locally on M5 Ultra Mac Studios. This technical partnership allows users to deploy the large language model on Apple Silicon hardware without relying on cloud-based inference services for execution.
According to CNBC, the configuration supports pairing two to four Mac Studios together to handle demanding tasks. These linked systems can perform coding operations and execute agentic workflows by distributing the computational load across the multiple high-memory machines.
Strategic Shift in AI Partnerships
This joint effort represents a notable change in the relationship between Meta and Apple. Tim Cook and Mark Zuckerberg have clashed publicly for years over platform policies and market dominance, creating a tense competitive environment between the two tech giants.
With John Ternus taking the reins at Apple, the rivals are finding common ground in artificial intelligence development. The collaboration suggests that hardware and software optimization for local AI is becoming a priority that transcends previous corporate disputes.
Technical Deployment Details
The Muse Spark Open model, which Meta AI Research published previously, is an 816B-parameter model. As Threat Times reported on October 15, 2026, Meta claimed this is the most capable US open-weight model with agentic desktop capabilities.
The earlier reporting noted that the weights enable multimodal agentic tasks on dual Mac Studios with 512GB memory each. The new announcement expands this capability, confirming that users can scale the setup by adding more units, up to four Mac Studios, to support the workload.
What should you do?
Security and technology professionals should assess their local compute infrastructure before attempting to deploy large open-weight models. Organizations considering this setup must verify that their network security policies allow for the necessary inter-machine communication between paired Mac Studios.
Administrators should also review the data privacy implications of running agentic workflows locally. While local execution keeps data off public clouds, the agentic nature of the model may still require access to internal repositories or APIs that need strict access controls.
Background: Large language models
Large language models are a specific type of neural network trained on text. Generative AI is the broader umbrella term for any system that creates new content, including images and code. Not all generative AI uses large language models, but most modern text generators do.
What to do and how to stay safe: Meta
- Verify that your organization's hardware inventory includes sufficient M5 Ultra Mac Studios with adequate memory to support multi-machine pairing.
- Review network firewall rules to ensure that inter-machine communication for distributed AI workloads is permitted and monitored.
- Assess the security posture of internal APIs and data repositories that the agentic model may attempt to access during local execution.
- Monitor for vendor updates regarding the Muse Spark Open model to ensure you are using the latest stable weights and documentation.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
How many Mac Studios can be linked to run Muse Spark Open?
Users can pair two to four M5 Ultra Mac Studios to run the model locally for coding and agentic work.
What is the size of the Muse Spark Open model?
The Muse Spark Open model has 816 billion parameters, making it a large-scale open-weight AI model.
Who is leading Apple in this new AI collaboration?
John Ternus is taking the reins at Apple, facilitating the common ground in artificial intelligence with Meta.
