Meta released the weights for Muse Spark Open on October 15. It is a sparse mixture-of-experts model with about 816 billion total parameters and about 42 billion active per token, using 256 experts per layer with 8 routed per token. It accepts text, images, audio and video, outputs text, and supports up to 1 million tokens of context.
Meta says the model was built for coding and agentic work, and that it is the first model in its class to run on a desktop setup: two Apple M5 Ultra Mac Studios using the latest MLX framework.
Hosted access is available at launch through Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter and Together AI. Weights are on Hugging Face, Amazon SageMaker AI, Google Cloud Model Garden and Microsoft Foundry. Hugging Face's write-up says the BF16 checkpoint is about 1.5 TB, and an NVFP4 checkpoint runs on a single B200 node.
The weights ship under the Muse Model License, which Meta describes as having limited commercial terms.