Meta has unveiled Muse Spark Open, which it calls the most capable US open-weight model to date. For Apple users, Meta says the model is optimized to run on the Mac Studio powered by the M5 Ultra. In a blog post, Meta explains that Muse Spark Open is natively multimodal and brings leading agentic capabilities to the desktop for the first time.
The model has 816B total parameters (42B active) and a 1M-token context window. Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning. Meta says it leverages Apple's latest version of MLX, an open source array framework created by Apple and optimized for the unified memory architecture of Apple silicon. Using that version, Meta says the model was optimized for the M5 Ultra Mac Studio, and that "together they're the best way to run a powerful model on your own desktop."
Apple's architecture lets the M5 Ultra's integrated GPU use up to 512 GB per Mac Studio, so two machines can hold an entire model. The chip's 1.2 TB/s of memory bandwidth is fast enough to read the small slice each token needs, so two Mac Studios behave like one supercomputer with a terabyte of unified memory. Meta tested agentic workflows running entirely on-device, including shipping real iOS changes to Ice Cubes, an open-source Mastodon client, and driving DaVinci Resolve end to end.
Muse Spark Open is available now through Hugging Face. Apple has also published an article on its Machine Learning Research website touting how MLX on Mac lets AI researchers and developers experiment locally in a private environment. The collaboration is a surprise given the icy history between the two companies. Meta AI chief Alexandr Wang told Axios the companies have been working together for several months, crediting former Apple executive Daniel Gross with playing a "really big role" in getting them to work well together.
Source: 9to5Mac · Summarized by HeadlinesBriefing