Meta Introduces Muse Spark Open: The Most Advanced Open-Weight Model in the U.S.
Meta has reaffirmed its dedication to open-source models by launching Muse Spark, a groundbreaking open-weight model known as Muse Spark Open. This impressive model boasts 800 billion parameters and has been specifically optimized for Apple Silicon, but it requires an extraordinary 1TB of RAM for local operation.
The company asserts that Muse Spark Open is “the most capable US open-weight model to date,” highlighting its competitive edge against two prominent Chinese models, GLM 5.3 and Kimi K3, particularly in agentic tasks and visual reasoning benchmarks.
Notably, this model is the first to be fine-tuned for local execution on Apple’s M5 Ultra-equipped Mac Studio systems. Meta points out that the M5 Ultra’s unified memory architecture allows its integrated GPU to utilize up to 512 GB per Mac Studio. combining two machines enables them to function collectively as a powerful supercomputer with a terabyte of unified memory.
Muse Spark Open — weights releasing today — runs great on new Mac Studios thanks to work with @johnternus and Apple. pic.twitter.com/U2wJNrKBOg
— Mark Zuckerberg (@finkd) October 15, 2026
As of now, Apple has not disclosed the pricing or release date for the 512GB Mac Studio, only indicating that it will be available “late October.” Any configuration capable of operating Muse Spark Open will demand a significant financial investment. The model can also run on four 256GB M5 Ultra Mac Studios, which together may cost around $38,000. This level of computational power is largely inaccessible for the average developer. for larger organizations that regularly utilize token budgets, acquiring several Mac Studios may present a worthwhile investment, especially if Meta’s model delivers the expected performance.
Muse Spark Open is now available for download from Hugging Face.

Access the original content here. The images featured in our article are sourced from this reference. We do not claim authorship; they have been used for informational purposes with appropriate credit to their original source.








