By No Code MBA · Facts checked
- Provider
- Meta
- Announcement date
- Status at last check
- Generally available
- Model family
- Muse Spark
- Capabilities
- Text, Image, Coding
What changed
Muse Spark Open is Meta's open-weight release in the Muse Spark family, a mixture-of-experts model with 816B total parameters and 42B active parameters and a 1M-token context window. Meta describes it as natively multimodal and aimed at coding and agentic work, including long-horizon software engineering and multi-step reasoning. Meta says it can run on two Apple M5 Ultra Mac Studios using MLX, and that custom Helion GPU kernels in SGLang give 1.03 to 3.95 times kernel-level speedups.
Access and availability
Weights are on Hugging Face at meta-models/Muse-Spark-Open-816B-A42B under the Muse Model License, and in Amazon SageMaker AI, Google Cloud Model Garden and Microsoft Foundry Models. Hosted inference is offered by Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter and Together AI, with support in vLLM, SGLang, TensorRT-LLM, Transformers, MLX, llama.cpp and Unsloth.
Official access documentationWhat you can build
Compare it with your current model on a bounded coding or multimodal agent task, tracking task completion and review effort on your own examples.
Limitations
The Muse Model License has limited commercial terms rather than a standard open-source license. The model needs very large memory to self-host, and Meta's full evaluation report was listed as available soon at release.