Meta releases 816-billion-parameter open-weight model Muse Spark Open, claims Mac compatibility but requires 1TB RAM

Meta released Muse Spark Open on October 15, an open-weight model with 816 billion parameters and a 1 million token context window. The company describes it as the "most capable" open-weight model in the United States, natively multimodal and specifically designed to run locally on Apple computers.

According to TechCrunch, while the model is built for Macs, it demands extreme hardware resources: at least 1 terabyte of RAM is required to run it, with optimal performance expected on Mac Studios equipped with the M5 Ultra chip. This high barrier means the model is currently accessible only to developers with top-tier workstation configurations, not general users.

Meta stated that the model was paired with its AI coding terminal, Muse Code, and deployed in various agentic workflows, including bug fixes, new features, and full builds for a Mastodon client. The company also demonstrated the model's ability to review and crop footage within DaVinci Resolve, a high-end video editing software.

It is important to note that claims regarding superior performance compared to other popular open-weight models are based on evaluation results released by Meta on Thursday. These are vendor self-reported benchmarks and have not yet been independently reproduced or verified by third parties.