Meta has unveiled Muse Spark Open, positioning it as the most capable open-weight AI model available in the United States. The model represents a significant leap in both scale and accessibility for developers working with large language models.
Muse Spark Open leverages 816 billion parameters and supports a 1 million token context window, enabling it to process and reason over extensive codebases and documents in a single pass. The model is natively multimodal, handling text, images, and structured data with equal fluency.
What sets Muse Spark Open apart is its optimization for Apple silicon. The model runs locally on Mac Studio systems equipped with at least 1 terabyte of RAM, bringing frontier AI capabilities to individual developers and small teams without requiring cloud infrastructure.
The model excels at complex coding tasks and agentic workflows, where AI systems must plan, execute, and adapt across multiple steps. Meta's evaluation results claim superior performance compared to other leading open-weight models across standard benchmarks.
The release comes as Meta continues its strategy of open-sourcing advanced AI models, contrasting with competitors who keep their most capable systems proprietary. This approach has made Meta's models foundational tools for researchers and startups building AI applications.