Meta has launched what it calls the “most capable” open-weight model in the United States. Dubbed Muse Spark Open, the model also has the distinction of having been designed to work with Macs.
The new model, which leverages 816 billion parameters and a 1 million token context window, claims to offer superior performance to other popular open-weight models, according to evaluation results released by Meta on Thursday.
Muse Spark Open is natively multimodal and built for complex coding tasks and agentic labor. It has also been designed to run locally on Apple’s computers. However, you will still need a lot of power to run the model locally, with a requirement of at least 1 terabyte of RAM, per Engadget. The model will run best via the Mac Studio that comes equipped with the M5 Ultra chip.
Still, if you can muster the memory, Muse Spark Open may prove a significant work horse for coding and other desktop tasks. Meta says that it paired the model with Muse Code, its AI coding terminal, and deployed it to various agentic workflows, including coding fixes for a Mastodon client that included “bug fixes, new features, [and] full build.”
The company says it also deployed it in the service of running DaVinci Resolve, the high-end video editing software, revealing that the model was capable of reviewing and cropping footage.
Open-weight models provide a number of distinct industry advantages that aren’t afforded by closed and proprietary models. Open-weight models can be significantly more cost-efficient and can also offer transparency and flexibility that closed source models don’t allow.
Meta is making the new model available now through a number of different API platforms.
Additional technical details from Meta AI Research (https://research.meta.ai/blog/muse-spark-open):
Muse Spark Open is a natively multimodal model with 816B total parameters (42B active) and a 1M-token context window. Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning.
Muse Spark Open is the first model in its class to run on your desktop using two Apple M5 Ultra Mac Studios with the latest MLX framework.
Available through API platforms at Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter and Together AI. Model weights on Amazon SageMaker AI, Google Cloud Model Garden, Hugging Face, and Microsoft Foundry Models.
Runs on SGLang, vLLM, TensorRT-LLM, Hugging Face Transformers, Apple MLX, llama.cpp, and Unsloth. Weights released under the Muse Model License with limited commercial terms.
Safety post-training and Meta Superintelligence Scaling Framework evaluations are described in the model card; evaluation report forthcoming.