Skip to content
Bramble

Technology

Meta releases Muse Spark Open model, built to run on Apple M5 Ultra Mac Studios

Meta released the open-weight Muse Spark Open model, which it calls the most capable US open-weight model. It was developed with Apple to run locally on Mac Studios.

2 outlets · 1L · 1C · 0R First reported Account updated
Image: The Verge
Image: TechCrunch

1 / 2

The story, neutrally told

Mixed · 2Meta announced Muse Spark Open on Thursday 15 October 2026, an open-weight AI model that it calls the "most capable US open-weight model to date". Left · 1Meta says it 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. Mixed · 2The model is natively multimodal and has 816 billion parameters and a 1 million token context window; Meta says it excels at long-horizon tasks, complex software engineering and multi-step reasoning.

Centre · 1TechCrunch says Meta's own evaluation results claim superior performance to other popular open-weight models, though these are Meta's figures. Left · 1Meta and Apple, typically rivals, worked together on the launch and are publicly backing each other; Meta CEO Mark Zuckerberg said it runs well on new Mac Studios thanks to work with Apple CEO John Ternus, who replied praising the efficiency of running such a model "right from your desk". Left · 1Meta chief AI officer Alexandr Wang told Axios, which noted the companies had collaborated for several months, "We’re excited for mommy and daddy to be playing nice again."

Left · 1Apple also published a technical post about MLX, Muse Spark Open and two other open models. Centre · 1Running it locally is demanding: TechCrunch, citing Engadget, says it needs at least 1 terabyte of RAM and runs best on a Mac Studio with the M5 Ultra chip. Centre · 1Meta says it paired the model with its AI coding terminal Muse Code and used it for tasks including fixes and new features for a Mastodon client and reviewing and cropping footage in DaVinci Resolve; it is available now through several API platforms.

Every sentence links to the reporting it rests on. The pill in front of each says where its sources sit: Left, Centre or Right when one side supplies at least half of them, Mixed when they are evenly split. The number is how many outlets it cites.

Left1 outlet

Framing
The Verge leads on the unusual Meta-Apple partnership between rivals, with executive quotes.
Emphasis
Zuckerberg, Ternus and Wang quotes and the collaboration.
Leaves out or plays down
Hardware requirements, benchmark claims and API availability.
Charged language
“mommy and daddy”
For example
“Meta and Apple, typically rivals, worked together on the launch and are publicly supporting each other.” — The Verge

Centre1 outlet

Framing
TechCrunch presents it as a product launch, focused on capabilities and practical use, with a note on hardware demands.
Emphasis
Specs, memory requirement, demo use cases and open-weight advantages.
Leaves out or plays down
The executive quotes and the rivalry angle; Apple's technical post.
Charged language
“significant work horse”
For example
“you will still need a lot of power to run the model locally, with a requirement of at least 1 terabyte of RAM” — TechCrunch

Right0 outlets

No right outlet in our sources has covered this story yet.

Ask about this story

Loading…