Meta launched Muse Spark Open, an 816-billion-parameter open-weight model with a million-token context window designed to run locally on Macs.
Meta launched Muse Spark Open on October 15, 2026, describing it as an open-weight model designed to run on Apple computers. The model has 816 billion parameters and a 1 million token context window. Meta said its evaluation results showed stronger performance than other popular open-weight models, though the article did not provide the underlying comparison results.
Muse Spark Open is multimodal and intended for complex coding and agent-based tasks. Running it locally requires substantial memory: TechCrunch, citing Engadget, reported a minimum of 1 terabyte of RAM. The article said the model is expected to run best on a Mac Studio equipped with Apple’s M5 Ultra chip, placing local use beyond the memory capacity of most ordinary computers.
Meta described pairing the model with Muse Code, its AI coding terminal, and using it in agent workflows. One example involved changes to a Mastodon client, including bug fixes, new features and a complete build. Meta also reported using the model with DaVinci Resolve to review and crop video footage. These examples show the range of tasks the company says it tested, rather than independently verified benchmark results.
Meta made Muse Spark Open available through several API platforms. The release combines an open-weight distribution with a design intended for local Mac execution, although the reported memory requirement is a significant practical constraint. Open weights can allow developers to inspect and adapt a model more than a closed service generally permits, but local deployment still depends on access to suitable hardware. The article did not specify API prices or provide a broader release timeline.