Meta and Apple have partnered to enable the 816-billion-parameter Muse Spark Open AI model to run locally on Apple M5 Ultra Mac Studios using the MLX framework.
Meta and Apple have announced a collaboration allowing developers to run Meta’s new Muse Spark Open AI model locally on Apple’s high-end desktop computers. The partnership marks a shift for the two companies, which have historically been rivals, as they work together to bring large-scale AI capabilities to desktop hardware.
Muse Spark Open is an open-weight artificial intelligence model with 816 billion total parameters, of which 42 billion are active at any given time. According to Meta, the model utilizes a mixture-of-experts architecture and features a 1-million-token context window. It is natively multimodal, meaning it can process text, images, video, and audio.
Meta describes the model as its "most capable US open-weight model to date." The company states that Muse Spark Open is designed for coding, complex software engineering, long-horizon tasks, and multi-step reasoning. In testing, Meta used the model with its Muse Code terminal to perform tasks such as making bug fixes and completing a full build of an open-source Mastodon client. The model was also tested with DaVinci Resolve, where it reviewed footage and identified sections for cropping.
The collaboration focuses on enabling the model to run on Apple’s M5 Ultra Mac Studios using Apple’s MLX machine learning framework. Meta says the model can run on a desktop setup using two Apple M5 Ultra Mac Studios, each configured with up to 512GB of unified memory. Together, these two machines provide 1TB of memory, allowing the model to operate without cloud infrastructure.
Apple published technical details on how to distribute the model across multiple Mac Studios using high-speed communication. Apple CEO John Ternus commented on the efficiency of the setup, stating, "Amazing to run a model of Muse Spark Open’s caliber and scale this efficiently, right from your desk. Mac Studio keeps pushing the boundaries of on-device AI!"
The local execution offers potential benefits in data security and cost savings by eliminating the need to offload tasks to cloud servers. However, the hardware requirements are significant. A single Mac Studio with the M5 Ultra chip and 256GB of memory costs at least $10,799, and a 512GB version is scheduled for release later this month. A four-machine configuration can cost over $40,000.
Meta CEO Mark Zuckerberg publicly supported the collaboration on X, writing, "Muse Spark Open -- weights releasing today -- runs great on new Mac Studios thanks to work with @johnternus and Apple." Meta Chief AI Officer Alexandr Wang noted the significance of the partnership, telling Axios, "We’re excited for mommy and daddy to be playing nice again."
Wang explained that many high-performance open-weight models are too large for standard personal computers, but the Mac Studio’s memory capacity makes local execution practical. He also credited Daniel Gross, an AI researcher and former Apple executive, as a factor facilitating the collaboration.
The partnership comes after years of tension between the two companies, particularly following Apple’s 2021 privacy changes that restricted ad tracking, which Meta estimated would cut its revenue by $10 billion. Neither company has confirmed if this collaboration signals a broader strategic alliance.
Muse Spark Open weights are available for download via platforms including Hugging Face, Amazon SageMaker AI, Google Cloud Model Garden, and Microsoft Foundry Models. The model is also accessible through API providers such as Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter, and Together AI. Apple has released updated MLX technical documentation to support the deployment of the model on Apple Silicon.
Sources
This article was drafted with AI assistance and checked against the sources above. Company claims are reported as claims. Cover image is AI-generated.
The day's checked news, by email
One short email a day with the top stories. Free, and you can unsubscribe in one click.
Or follow for the latest news:
No comments yet. Start the discussion.
Microsoft introduced new security features to monitor AI agents, warning that autonomous systems can pose insider risks even without being hacked. The company urges separating intelligence supply from control authority.
The Department of Homeland Security has ordered federal civilian agencies to remove DeepSeek AI products from IT systems and barred future procurements, citing cybersecurity risks.