Meta Launches Muse Spark Open for Local Multimodal Use

On October 16, 2026, Meta released Muse Spark Open, a multimodal model built with 816 billion parameters. While the total scale of the system is vast, the architecture is designed to use 42 billion active parameters per request, allowing it to run on two Apple M5 Ultra Mac Studios. With a one-million-token context window, the model handles text, image, video, and audio processing within a single framework. It is available for local commercial use under Meta's Muse license, which enables businesses and developers to integrate these multimodal features directly into their own applications.

The development of Muse Spark Open follows a broader shift toward multimodal systems in artificial intelligence. Historically, AI models were limited by their focus on a single medium, such as text or images. As computing power and algorithmic efficiency have improved, researchers have moved toward models that can process and understand several forms of data at once. This model reflects that transition, offering a more integrated approach to how software interprets information.

In practical terms, the model is expected to give developers the tools necessary to build more interactive applications. The project was managed by Meta's AI research team, who handled the design, training, and optimization of the system. A partnership with hardware manufacturers like Apple ensures that the software runs efficiently on high-end local machines. By making this technology available for local installation, Meta allows businesses to utilize AI capabilities without relying on cloud-based services.

The move toward local use also carries implications for how businesses develop software tailored to specific regional or organizational needs. Because the model can manage diverse data types, it may find applications in sectors ranging from education and healthcare to entertainment. As developers explore the features of Muse Spark Open, the system has the potential to support new creative directions and technical adjustments across the industry.

This release represents a specific stage in AI development, and Meta intends to continue refining the model based on technological changes and user feedback. By providing broader access to these tools, the release allows a wider variety of organizations to work with sophisticated AI systems. As the field continues to change, Muse Spark Open may serve as a foundation for future work in multimodal AI applications.