Meta is releasing Muse Spark Open as an open-weight AI model designed to run on a powerful desktop setup built from Apple hardware. The announcement also points to an unusual moment of cooperation between Meta and Apple, companies that are typically rivals.
A large model aimed at desktop use
Meta describes Muse Spark Open as a natively multimodal model with 816 billion total parameters. The company says it excels at long-horizon tasks, complex software engineering, and multi-step reasoning.
Meta also calls it the “most capable US open-weight model to date.” It says Muse Spark Open is the first model in its class that can run on a desktop using two Apple M5 Ultra Mac Studios with the latest MLX framework.
That setup is notable because it frames the model as something users can run locally on desktop computers, rather than only access through a remote service. The announcement does not provide further details about the hardware configuration or what performance users should expect.
Meta and Apple publicly back the work
The companies worked together on the launch and have been collaborating over several months, according to Axios. Their leaders also publicly acknowledged the joint effort.
Meta CEO Mark Zuckerberg said, “Muse Spark Open -- weights releasing today -- runs great on new Mac Studios thanks to work with @johnternus and Apple,” in a post on X. Apple CEO John Ternus replied, “Congrats to the team!” and said it was amazing to run a model of Muse Spark Open’s caliber and scale efficiently from a desk.
Ternus also said Mac Studio keeps pushing the boundaries of on-device AI. Alexandr Wang, Meta’s chief AI officer, described the companies’ renewed cooperation to Axios with the remark, “We’re excited for mommy and daddy to be playing nice again.”
MLX provides the Apple Silicon framework
Apple published a technical post about MLX, describing it as an “array framework” for machine-learning tasks on Apple Silicon. The post covers Muse Spark Open alongside two other open models.
In this announcement, MLX is the link between Meta’s model and Apple’s hardware. Meta says the model can run using the framework on two Mac Studios, while Apple’s post presents MLX in the context of multiple open models.
Together, the announcements show both companies highlighting local AI computation on Apple Silicon. Meta brings the open-weight model; Apple contributes a framework and hardware platform that the companies say can run it at this scale.
What the announcement establishes
The launch establishes a concrete claim about Muse Spark Open’s desktop setup, as well as a public endorsement of the collaboration by executives at both companies. It also positions the model around capabilities Meta says are important for extended tasks, software engineering, and reasoning across multiple steps.
The source provides no benchmark results or comparative measurements, so the model’s performance claims remain Meta’s description. Still, the collaboration signals that Meta and Apple are willing to work together on making a large open-weight model run on Apple desktop hardware.