AIHere

Meta releases open-weight Muse Spark Open, tuned with Apple for Mac Studio

Meta released its 816-billion-parameter open-weight Muse Spark Open model on October 15, 2026, saying it can run on two Apple M5 Ultra Mac Studios.

AIHere News Desk· 2 min read🇺🇸 United States
Illustration of two desktop computers linked together to run a large AI model.
Photo: KKPCW(Kyu3) / Wikimedia Commons, CC BY-SA 4.0

Meta unveiled Muse Spark Open on October 15, 2026, publishing the model's weights the same day. The company calls it the most capable open-weight model yet produced in the United States. The launch was notable for a public partnership with Apple, a company Meta usually competes with.

According to Meta, Muse Spark Open is the first model in its class that can run on a desktop setup. The company said it runs locally on two Apple M5 Ultra Mac Studios using the latest version of Apple's MLX framework.

A large multimodal model

Meta describes the model as natively multimodal, with 816 billion total parameters. It claims the system performs especially well on long-horizon tasks, complex software engineering and multi-step reasoning. Those performance claims come from Meta and have not yet been independently verified.

Apple published a technical post the same day about MLX, its array framework for machine learning on Apple Silicon. The post covered Muse Spark Open along with two other open models.

Executives trade praise

Meta CEO Mark Zuckerberg wrote on X that the model runs very well on the new Mac Studios thanks to work with Apple and its chief executive, John Ternus. Ternus, Apple's new CEO, replied with congratulations, calling it remarkable to run a model of this scale so efficiently from a desk. Meta chief AI officer Alexandr Wang joked about the two firms getting along again, saying, "We're excited for mommy and daddy to be playing nice again." The companies had reportedly been collaborating for several months.

The tie-up highlights a push to run frontier-scale models on personal hardware rather than only in the cloud. For Meta, open weights help build developer adoption; for Apple, the demonstration positions the Mac Studio as a serious machine for on-device AI. Still, the need for two high-end workstations means the setup will appeal mainly to well-equipped developers and researchers.

#Meta#Apple#open-weight models#Muse Spark Open#MLX#on-device AI

Comments

Be respectful. Comments are checked automatically; spam, insults and personal data are removed. Questions may be answered by AI.

No comments yet. Start the discussion or ask a question.

Related