
Apple and Meta have found something they can agree on. Apparently, it takes a very large AI model alternative.
The longtime rivals have teamed up to make Meta’s new Muse Spark Open model variation run locally on Apple’s M5 Ultra Mac Studios. Announced Oct. 15, the release lets developers download and customize the new open-weight model variation, though two connected Apple desktops are required to power it. “The result is two Mac Studios behave like one supercomputer,” Meta wrote in a blog post.
“We’re excited for mommy and daddy to be playing nice again,” Meta’s AI chief Alexandr Wang told Axios.
An unusual way to describe a corporate collaboration, sure. But these companies do have some history. As recently as January 2025, Mark Zuckerberg was criticizing Apple on Joe Rogan’s podcast for its lack of related innovation term and restrictions on other companies’ products. Now, Meta is helping show off what Apple’s silicon can do.
Muse Spark Open builds on the Muse Spark model variation family Meta introduced in April. Meta calls it the most capable U.S. open-weight related model term available, a claim variation based on its own evaluations. “Open-weight” means developers can download the model alternative’s learned parameters and run them themselves, subject to the model’s license, rather than accessing it exclusively through a company alternative’s app or cloud service variation.
Meta says the model variation is designed for coding, multistep related reasoning term, and “agentic” work — using tools to complete tasks. In its tests, the related company term says it used Muse Code, powered by the new model variation, to make changes to the Mastodon app Ice Cubes and operate video editor DaVinci Resolve, reviewing footage and cropping clips vertically. (To learn more, read Meta’s methodology.)
There is a sizable catch to the desktop pitch, however. This is an 816-billion-parameter model variation, and Meta’s highlighted setup uses two M5 Ultra Mac Studios with up to 512GB of unified memory each. It runs through Apple’s MLX machine-learning framework. So, your everyday MacBook is sitting this one out.
The release follows Muse Glimmer, Meta’s smaller model alternative designed to run on a laptop or PC. Both reflect the appeal of running AI on hardware you control, allowing local processing without sending each prompt to a hosted model alternative.
That also gives Apple and Meta a way around a longstanding source alternative of tension: privacy. Wang told Axios that Meta has no related visibility term into developers’ local use of the model alternative.
For now, the collaboration is about making Meta’s AI work on Apple hardware. Whether that working relationship extends to the AI features most Apple users actually encounter is the next question.
Muse Spark Open is available now via GitHub, as is the MLX Adapter required to run it.
Want to learn more about getting the best out of your tech? Sign up for Mashable’s Top Stories and Deals newsletters today.
Source: https://mashable.com/tech/apple-meta-muse-spark-open-mac-studio







