Meta has released Muse Spark Open, an open-weight model it calls the most capable open-weight model in the United States and one explicitly designed to run on Apple Macs. The model has 816 billion parameters and a 1 million token context window; Meta‘s evaluation results claim superior performance to other popular open-weight models. It is natively multimodal and aimed at complex coding tasks and agentic work.

Running locally on Apple hardware is a headline feature, but it carries a large memory requirement: at least 1 terabyte of RAM, per Engadget, with the best experience on a Mac Studio equipped with the M5 Ultra chip. Meta says it paired the model with its Muse Code AI coding terminal and used it in agentic workflows, including applying bug fixes, new features, and full builds to a Mastodon client. The company also used the model to run DaVinci Resolve, where it could review and crop footage.

Meta emphasizes the general advantages of open-weight models over closed ones, including cost efficiency, transparency, and flexibility. Muse Spark Open is available now through multiple API platforms.