Meta Launches Muse Spark Open For Apple Silicon Macs
Meta’s new open-weight AI model targets coding and other complex tasks, but running it locally requires at least 1TB of memory.
Meta has launched Muse Spark Open, an open-weight AI model optimized to run locally on Apple Silicon Macs. Meta calls it the most capable US open-weight model to date, a claim reported by TechCrunch, The Verge and Engadget. The desktop setup is substantial: Engadget says running the model locally requires 1TB of memory.
Table of Contents
What Meta released
Muse Spark Open is a multimodal model intended for complex software engineering, extended tasks and multi-step reasoning, according to The Verge. TechCrunch reports that it has 816 billion parameters and a context window of 1 million tokens. Engadget describes it as an 800-billion-parameter model.
Meta’s own evaluation results suggest it outperforms other open-weight models, TechCrunch reports. Engadget says Meta’s benchmarks specifically show an advantage over GLM 5.3 and Kimi K3 on agentic tasks and visual reasoning. Those performance comparisons come from Meta’s released results.
Why running it on a Mac takes two machines
The Verge reports that Meta describes Muse Spark Open as the first model in its class able to run on a desktop using two M5 Ultra Mac Studios and the latest MLX framework. Engadget explains that each Mac Studio in that configuration can provide up to 512GB of unified memory, allowing the pair to hold the model.
Engadget says four 256GB M5 Ultra Mac Studios are another option, at a cost of roughly $38,000. It also reports that Apple has not announced a price or specific ship date for the 512GB Mac Studio, saying only that it will be available in late October. Meta and Apple worked together on the launch, according to The Verge, which reports that their leaders publicly praised the collaboration.
Muse Spark Open is available to download from Hugging Face, Engadget reports. TechCrunch says Meta is also making the model available through multiple API platforms.
Potential use for software teams
TechCrunch reports that Meta paired the model with its Muse Code coding terminal and used it in workflows that included fixes, new features and a full build for a Mastodon client. Meta also demonstrated it reviewing and cropping footage in DaVinci Resolve, according to TechCrunch.
For teams considering local AI for coding or other desktop work, the hardware requirement is a significant constraint. Engadget says the cost puts a local setup beyond the reach of a typical individual coder, while suggesting it could interest larger organizations seeking an alternative to recurring token spending—if the model performs as Meta claims.
Sources
This story was compiled by AI from the reports below. Read the originals for the full details.