Meta and Apple team up to make Muse AI models run better on Macs
News Source
β’2026-10-15T14:15:36-04:00
π° What Happened
Meta has announced a new AI model called Muse Spark Open. The company says it is the most capable open-weight model built in the United States so far. Meta also says it is the first model of its kind that can run on a desktop, using two Apple M5 Ultra Mac Studios and Apple's latest MLX software. Apple published a technical post about the setup. The model has 816 billion total parameters and is built to handle long tasks, complex coding and multi-step reasoning. Both companies' leaders publicly supported each other's work.
π The Backstory
Meta and Apple have often competed, so the partnership is notable. Meta's chief AI officer joked that it was good for mommy and daddy to be playing nice again. An open-weight model is one whose core is released so others can download and run it. Running a huge model on desktop hardware, instead of in a remote data centre, is a step toward more private and local AI that people can control themselves.
π― Why It Matters
If powerful models can run on a personal computer, users depend less on big cloud servers. That could change the cost of AI, improve privacy, and shift who controls the technology. It also shows that even fierce rivals may work together when the stakes are high. It also signals that the race for AI may increasingly be about hardware people own, not just cloud services.
Meta has announced a new AI model called Muse Spark Open. The company says it is the most capable open-weight model built in the United States so far. Meta also says it is the first model of its kind that can run on a desktop, using two Apple M5 Ultra Mac Studios and Apple's latest MLX software. Apple published a technical post about the setup. The model has 816 billion total parameters and is built to handle long tasks, complex coding and multi-step reasoning. Both companies' leaders publicly supported each other's work.
Meta and Apple have often competed, so the partnership is notable. Meta's chief AI officer joked that it was good for mommy and daddy to be playing nice again. An open-weight model is one whose core is released so others can download and run it. Running a huge model on desktop hardware, instead of in a remote data centre, is a step toward more private and local AI that people can control themselves.
If powerful models can run on a personal computer, users depend less on big cloud servers. That could change the cost of AI, improve privacy, and shift who controls the technology. It also shows that even fierce rivals may work together when the stakes are high. It also signals that the race for AI may increasingly be about hardware people own, not just cloud services.