Meta has released Muse Spark Open, an open-weight AI model featuring 816 billion parameters and a 1 million token context window. The company claims the model offers superior performance compared to existing open-weight alternatives.
The model is natively multimodal and optimized for complex coding tasks and agentic workflows. Meta has integrated it with its Muse Code terminal and demonstrated its capability to handle tasks such as bug fixes and video editing in DaVinci Resolve.
Designed to run locally on Apple hardware, the model requires at least 1 terabyte of RAM for operation. Meta recommends using a Mac Studio equipped with an M5 Ultra chip for optimal performance.
The new model is currently available through various API platforms. Meta highlights that its open-weight nature provides users with increased transparency, flexibility, and cost-efficiency compared to proprietary models.
