Meta introduced Muse Spark Open, an open-weight AI model with 816 billion parameters and a 1 million token context window, designed to run on Macs. The company announced the model on Thursday, highlighting its capability to perform complex coding and agentic tasks. Muse Spark Open requires at least 1 terabyte of RAM to run locally, making it best suited for Apple’s Mac Studio equipped with the M5 Ultra chip, according to techcrunch.com.
The model is multimodal and built to handle sophisticated workflows, including coding and video editing. Meta paired Muse Spark Open with Muse Code, its AI coding terminal, and demonstrated its use in fixing bugs, adding features, and building a Mastodon client. Additionally, the model was deployed to assist with DaVinci Resolve video editing software by reviewing and cropping footage, showcasing its versatility in desktop applications, techcrunch.com reported.
Open-weight models like Muse Spark Open differ from closed proprietary models by offering greater flexibility and transparency, which can accelerate innovation in AI development. Meta’s release positions it against other large open models, emphasizing its scale and native support for Macs. The model’s large parameter count and extensive context window suggest strong performance in coding and multimodal tasks, areas increasingly important in AI-driven workflows, according to techcrunch.com.
Muse Spark Open’s launch marks a notable step in AI models optimized for Apple hardware, targeting professional users with demanding workloads. The model’s deployment alongside Muse Code and integration with applications like DaVinci Resolve underline its practical utility. Meta’s announcement on October 15 confirms the model is available for use on Macs capable of meeting its high memory requirements, as detailed by techcrunch.com.