Meta and Apple collaborate on Muse AI models for Apple Silicon Macs

In a notable collaboration between two tech giants, Meta and Apple have joined forces to optimize new AI models for Apple's hardware. This partnership, centered on Meta's newly announced Muse Spark Open model, signals a push to make powerful, open-weight AI models viable for local, on-device execution, potentially reshaping the accessibility frontier for developers and creators.

What is Muse Spark Open and why does it matter?

On October 15, 2026, Meta announced the release of Muse Spark Open, describing it as the "most capable US open-weight model to date," according to a report by The Verge. The model is natively multimodal and contains 816 billion parameters, with Meta stating it excels at long-horizon tasks, complex software engineering, and multi-step reasoning. Crucially, Meta claims it is the first model in its class designed to run locally on a desktop using two Apple M5 Ultra Mac Studios. This move represents a significant step toward democratizing access to frontier-scale AI models by enabling them to operate outside of cloud APIs, offering greater control, privacy, and potentially lower latency for users with the requisite hardware.

How did Apple and Meta collaborate on this launch?

The announcement was marked by public support from the highest levels of both companies, a notable event given their typical rivalry. Meta CEO Mark Zuckerberg stated on X that the model "runs great on new Mac Studios thanks to work with @johnternus and Apple." Apple's new CEO, John Ternus, replied, congratulating the team and highlighting the efficiency of running a model of "Muse Spark Open’s caliber and scale this efficiently, right from your desk." Meta's chief AI officer, Alexandr Wang, commented to Axios, "We’re excited for mommy and daddy to be playing nice again," indicating a collaboration that had been ongoing for several months. Apple also published a new technical post about its MLX framework in relation to the launch.

What is Apple's MLX framework and its role?

Central to this collaboration is Apple's MLX, an array framework designed for machine learning tasks on Apple Silicon. According to the reporting, MLX is the underlying technology that enables Muse Spark Open to run efficiently on the M5 Ultra Mac Studios. Apple's technical post, released alongside the model announcement, detailed MLX's role alongside Muse Spark Open and two other open models. This framework is Apple's answer to providing developers with optimized tools for deploying AI models on its hardware, creating a more integrated and performant local AI stack that competes with cloud-based inference services.

What does this mean for AI developers and creators?

For developers and researchers, this collaboration lowers the barrier to experimenting with and deploying a large, multimodal model. Instead of relying solely on cloud credits or API rate limits, those with high-end Apple hardware can run inference locally. This enables more extensive testing, fine-tuning, and integration into applications where data privacy or network independence is paramount. For creators, especially in fields like software engineering, multimedia, and complex content generation, it opens a path to using a powerful AI assistant directly on their workstation, integrated into their local workflow without sending data to external servers.

How does this relate to the AI tool ecosystem in Grafy?

While Grafy does not locally host models of this scale, its architecture is built to integrate and leverage a diverse ecosystem of cloud models. The development highlighted by Meta and Apple—making powerful models more accessible and efficient—parallels Grafy's core philosophy of providing a unified workspace where users can access the right model for the task. The Graph app, for instance, allows users to visually branch through prompts and outputs across different models, a workflow that becomes even more powerful as the model landscape diversifies between cloud and local options. For users interested in the technical specifics of model frameworks and deployment, exploring the topic further is possible within Grafy's Reader app by analyzing related PDFs and technical papers, or by discussing the implications directly with various AI models in Chat.

What are the broader implications of this partnership?

This partnership between Meta and Apple challenges the prevailing assumption that the largest AI models are exclusively the domain of cloud infrastructure. By co-engineering a model for Apple's high-end consumer hardware, they are creating a new category: the professional-grade, locally-runnable frontier model. This could accelerate a trend toward hybrid AI workflows, where initial prototyping or sensitive tasks are handled on-device, while larger-scale batch jobs are sent to the cloud. It also strengthens the position of open-weight models in the ecosystem, providing a tangible, high-performance alternative to closed API models from other major players.

Frequently asked questions

### What hardware is required to run Muse Spark Open locally?

According to Meta, the model is designed to run using two Apple M5 Ultra Mac Studios with the MLX framework. This represents a high-end, professional desktop setup, indicating that local execution is currently targeted at developers and studios with significant computational resources.

### Is Muse Spark Open truly "open"?

Meta describes it as an "open-weight" model, which typically means the model weights are publicly released for others to use and study. This differs from "open source," which usually includes training code and data. The announcement focuses on the release of the weights, aligning with Meta's previous approach with models like Llama.

### How does this affect the use of AI cloud services?

This development offers an alternative to cloud APIs for users with the necessary hardware, providing benefits in privacy, cost control for high-volume use, and latency. It is unlikely to replace cloud services for most users but creates a viable on-premise option for specific professional and development use cases.

---

Source: The Verge — AI — Meta and Apple team up to make Muse AI models run better on Macs