Meta has introduced Muse Spark Open, an open-weight AI model that the company describes as one of the most capable models developed in the United States. It is also designed to run on Apple Mac computers.

 

The model features 816 billion parameters and a 1-million-token context window, enabling it to handle complex coding tasks and agentic workflows. According to evaluation results released by Meta, it aims to deliver competitive performance against other leading open-weight AI models.

 

Muse Spark Open is natively multimodal, meaning it can process different types of information. It is built to support advanced programming, automate desktop tasks, and power AI agents that can carry out multi-step operations.

 

However, running the model locally requires substantial computing resources. According to Engadget, users need at least 1TB of RAM for local deployment, with a Mac Studio featuring Apple's M5 Ultra chip expected to provide a suitable setup.

 

Meta has demonstrated the model's coding capabilities by integrating it with Muse Code, its AI coding terminal. In one example, the system worked on a Mastodon client, handling bug fixes, adding features, and completing a full build.

 

The company has also tested Muse Spark Open with DaVinci Resolve, professional video-editing software. The model reportedly demonstrated the ability to review video footage and crop clips, highlighting its potential for creative and desktop-based workflows.

 

Open-weight AI models offer several advantages over fully closed systems. They can give developers greater flexibility to customize deployments, improve transparency into model behavior, and potentially reduce costs, depending on the hardware and infrastructure required.

 

Meta is making Muse Spark Open available through several API platforms, giving developers opportunities to explore its capabilities for coding, automation, and other AI-powered applications.