Apple and Meta are working together to bring Muse Spark Open, a powerful open-weight artificial intelligence (AI) model, to high-end Mac computers. The collaboration marks an unusual partnership between two technology giants that have long disagreed over privacy, digital platforms, and the future of computing.
Meta is introducing Muse Spark Open as an open-weight AI model with 816 billion parameters. The company describes it as the most powerful open-weight AI model developed in the United States currently available.
The announcement highlights growing efforts to run advanced AI models directly on personal computers. Instead of depending entirely on cloud-based services, users and developers could run demanding AI workloads locally with the right hardware.
Muse Spark Open to Run on Apple Silicon Macs
As part of the collaboration, Meta has worked with Apple to enable Muse Spark Open to run locally across two or four connected Mac Studios powered by Apple’s M5 Ultra chips.
The setup relies on Apple’s MLX framework, which is designed to support machine learning workloads on Apple silicon. This allows developers to explore running large AI models on Apple’s hardware rather than relying exclusively on remote servers.
Running an AI model with 816 billion parameters requires substantial computing power and memory. Connecting multiple high-performance Mac systems provides a way to distribute the workload across available hardware resources.
The demonstration also highlights how Apple’s desktop computers could become more useful for researchers, developers, and businesses working with advanced AI technologies.
Free Download and Commercial Use
Meta plans to make Muse Spark Open available as a free download. Its licensing terms will also permit commercial use, potentially allowing businesses and developers to experiment with the model and build their own AI-powered applications.
In addition to local installation, the model will be accessible through cloud providers and developer tools. This approach could give users more flexibility in choosing how they deploy AI systems, depending on their computing resources and technical requirements.
Open-weight models allow developers to access a model’s learned weights, subject to the applicable license. However, open-weight availability does not necessarily mean that every part of a model’s training data or development process is publicly accessible.
For businesses, access to a model that can be deployed locally may offer greater control over infrastructure and data handling. Actual privacy benefits will depend on how the model is installed, configured, and used.
Why the Apple and Meta Partnership Matters
The collaboration is notable because Apple and Meta have spent years competing over privacy policies, app distribution, and their broader visions for digital technology.
Apple has promoted privacy and on-device processing as important features of its ecosystem. Meta, meanwhile, has invested heavily in artificial intelligence and open-weight models to expand access to its technology.
Bringing a large AI model to high-end Macs demonstrates a potential area of common interest despite their wider differences. Both companies can benefit from advances that make powerful AI systems more accessible to developers and other users.
The development could also strengthen the role of Apple silicon in AI computing. If local AI workloads become more practical on desktop hardware, developers may have additional options beyond traditional cloud infrastructure and specialized data-centre systems.
However, running such a large model locally is not the same as running a lightweight chatbot on a standard laptop. The demonstrated setup requires multiple high-end Mac systems, meaning the hardware cost and technical requirements could limit access for everyday users.
What It Means for the Future of AI
The arrival of Muse Spark Open on connected Mac Studios reflects a wider shift towards flexible AI deployment. Companies are increasingly exploring ways to run models on local devices, private infrastructure, and cloud platforms.
For developers, the availability of a free model with commercial-use permissions could create opportunities to test applications, develop new tools, and experiment with advanced AI capabilities.
For Apple, the demonstration offers another example of how its high-performance hardware can support demanding machine learning tasks. For Meta, collaboration with Apple could help extend the reach of its open-weight AI technology.
The next major question will be how accessible and practical these systems become outside specialised demonstrations. Performance, memory requirements, software compatibility, licensing conditions, and hardware costs will all influence wider adoption.
For now, the Apple and Meta collaboration signals an important development in the effort to bring powerful AI models beyond cloud-based platforms and onto local computing systems. It also shows how companies that compete in many areas can still work together when their technological goals align.