A large open-weight model for local workflows
Meta has introduced Muse Spark Open, an open-weight AI model designed to run on Apple computers. The company says it has 816 billion parameters and a context window of up to one million tokens.
Meta describes the model as natively multimodal and suited to demanding coding tasks and AI-agent workflows. Its evaluations, published by the company, indicate strong performance compared with other open-weight models.
Running it locally requires substantial hardware: the stated minimum is 1 terabyte of RAM. Meta says it works best on a Mac Studio equipped with an M5 Ultra chip. The model is also available through several API platforms.
Meta has tested Muse Spark Open with its Muse Code coding terminal, using it for tasks such as fixing bugs, adding features and building software. The company also says it has applied the model to other desktop workflows.
Open-weight models can give developers greater flexibility and transparency than closed systems, while potentially reducing costs. As powerful models become more accessible for local use, they could bring increasingly capable AI tools directly into everyday computing workflows.
