Meta releases Muse Spark Open, an 816B open-weight model that runs on two Mac Studios

Meta has published the weights of Muse Spark Open, which it calls the most capable US open-weight model so far. The headline twist is where it can run: Meta worked with Apple so that the model fits on a pair of desktop Mac Studio computers instead of a data-centre rack.

The model

According to Meta’s announcement, Muse Spark Open is natively multimodal and uses a mixture-of-experts design with 816 billion parameters, of which 42 billion are active for each token. It has a 1-million-token context window and is tuned for coding, agentic work, long-horizon tasks and multi-step reasoning. It arrives days after Meta’s closed Muse Wonder models went on sale through its API.

Why two Mac Studios are enough

The M5 Ultra version of Mac Studio can give its integrated GPU up to 512 GB of unified memory, so two machines together hold the whole model — roughly a terabyte of memory between them. Meta says the chip’s 1.2 TB/s memory bandwidth is fast enough because a mixture-of-experts model only reads a small slice of its weights per token. CNBC reports that customers can link two to four Mac Studios, and that Apple published documentation on how its MLX framework splits open models across several machines.

Meta says it tested on-device agent workflows with its Muse Code tool, including shipping changes to the open-source Mastodon app Ice Cubes and driving DaVinci Resolve to pick and crop video clips.

Where to get it

• Weights: Hugging Face, Amazon SageMaker AI, Google Cloud Model Garden and Microsoft Foundry Models.
• Hosted APIs: Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter and Together AI.
• Engines: SGLang, vLLM, TensorRT-LLM, Transformers, MLX, llama.cpp and Unsloth, with optimisations for AMD, Apple, Intel and NVIDIA hardware.

The licence is the Muse Model License with limited commercial terms, so check it before building a product on the model.

For developers, the bigger story is that a frontier-class open model is now practical on hardware that sits on a desk — expensive hardware, but no data centre required.