Meta announced the launch of Muse Spark Open, which the company describes as the most capable open-weight AI model available in the United States. The model is explicitly built to operate on Apple Macintosh computers, marking a rare focus on native Mac compatibility among large-scale language models.
Muse Spark Open contains 816 billion parameters and supports a context window of one million tokens, according to Meta’s evaluation released on Thursday. The model is natively multimodal and engineered for complex coding assignments and agentic labor, positioning it for tasks that blend natural-language understanding with programmatic execution.
Running the model locally demands substantial hardware, with Meta citing a minimum of one terabyte of RAM as a prerequisite. Engadget reports that performance peaks on the Mac Studio equipped with Apple’s M5 Ultra chip, suggesting that only the most powerful Mac configurations can fully exploit Muse Spark Open’s capabilities.
Meta paired Muse Spark Open with its Muse Code terminal, an AI-driven coding environment. The combined system was deployed to address agentic workflows, including fixing bugs, adding new features, and delivering a full build for a Mastodon client, demonstrating the model’s utility in real-world software development cycles.
The model was also integrated with DaVinci Resolve, a high-end video editing suite, where it performed tasks such as reviewing footage and cropping clips. This application illustrates Muse Spark Open’s capacity to support creative workflows beyond code, extending its relevance to media production professional environments.
Open-weight models like Muse Spark Open offer industry advantages not found in closed, proprietary systems. Meta highlights lower cost, greater transparency, and flexibility as key benefits, allowing organizations to customize and audit the model without vendor lock-in, potentially reshaping AI future deployment strategies across enterprises.