Meta and Apple revealed a new collaboration that will enable developers to easily run Meta's new open-weight AI model on Apple's high-end desktop computers, CNBC reported.[1] The partnership lets developers run a frontier-class model locally instead of in the cloud, which the companies said adds security and reduces costs.[1][2] This move signals a shift in Silicon Valley alliances, as Apple and Meta have avoided public partnerships since 2017, CNBC reported.[1]

Muse Spark Open is an 816-billion-parameter model using a mixture-of-experts architecture, which reduces compute at inference by keeping only a fraction of its parameters active, Apple Machine Learning Research said.[2][1] The model is natively multimodal and built for complex coding and agentic labor, TechCrunch reported.[3]

Customers will be able to pair between two and four Mac Studio PCs to use Muse Spark, CNBC reported.[1] Apple reported that on a cluster of four Mac Studios, Muse Spark Open generates over 36 tokens per second at a context size of 8,192 tokens.[2] The M5 Ultra chip provides 1.2 TB/s of unified memory bandwidth, compared to 819 GB/s for the M3 Ultra, Apple Machine Learning Research said.[2]

The hardware required for this setup is expensive; a Mac Studio with an M5 Ultra chip and 256GB of memory costs at least $10,799, CNBC reported.[1] A version with 512GB of RAM is scheduled for release later this month, CNBC reported.[1] Meta is also opening a private API preview to select users, Meta AI said.[4]