Meta Unveils Muse Spark Open, Dubbed “America’s Most Powerful Open-Weight Model,” Optimized for Mac Studio
Meta has released Muse Spark Open, an open-weight model it claims is the most capable open-weight model from the U.S. to date. It is a native multimodal model with a total of 816 billion parameters (420 billion active), a 1-million-token context window, and is optimized for Mac Studios equipped with the M5 Ultra and Apple's MLX framework, making it suitable for coding and agent tasks. Meta states it leads on multiple benchmarks like JobBench, though it still lags behind paid models.
Meta has launched Muse Spark Open, an open-weight model it claims is the most capable open-weight model from the U.S. to date. The model is natively multimodal, with a total of 816 billion parameters and 42 billion active parameters, a context window of 1 million tokens, and optimizations for Mac Studio devices equipped with the M5 Ultra. Leveraging Apple's latest MLX framework, it is well-suited for programming and agent tasks.
Meta states that it outperforms Qwen 3.8, GLM 5.3, and Kimi K3 on multiple benchmarks, including JobBench, though it still falls short of paid models.
