Meta released Muse Spark Open on October 15, 2026, making the AI model’s weights available for download and enabling local operation on Apple’s M5 Ultra Mac Studio computers. The release expands developers’ deployment options beyond hosted services, although the Mac setup described by Meta uses two high-memory desktops rather than a single consumer computer.
Key Takeaways
- Meta released the model on October 15, with downloadable weights and hosted access.
- The highlighted Apple configuration runs across two M5 Ultra Mac Studios using MLX.
- Commercial use is governed by the Muse Model License, not unrestricted open-source terms.
Table of Contents
- Muse Spark Open brings a large model to the desktop
- Apple support sits within a broader developer rollout
- Downloadable weights come with license conditions
- Conclusion
Muse Spark Open brings a large model to the desktop
Meta designed the model for coding and agentic work: tasks in which software plans steps, uses tools and checks results. Its multimodal design supports work across different types of input rather than text alone.
In its Muse Spark Open release announcement, Meta describes local operation on two M5 Ultra Mac Studios using the latest MLX framework. Each machine can provide up to 512 GB of unified memory, allowing the pair to hold the model while sharing computation.
That hardware requirement is an important qualification to the desktop-AI pitch. The announcement demonstrates a local deployment option on a substantial workstation setup; it does not establish that the full model runs on every Mac.
| Item | Details |
|---|---|
| Model size | 816 billion total parameters |
| Active parameters | 42 billion per token |
| Context window | 1 million tokens |
| Mac memory bandwidth | 1.2 TB/s per M5 Ultra chip, as specified by Meta |
Apple support sits within a broader developer rollout
Meta’s developer announcement identifies Apple among the companies it worked with to provide launch-day software support. The confirmed development is a Meta model release with Apple hardware and framework support, rather than a separately established joint Apple–Meta product announcement.
The rollout is not exclusive to Apple. Meta lists support for inference software including SGLang, vLLM, TensorRT-LLM, Hugging Face Transformers and llama.cpp, with optimizations for AMD, Apple, Intel and NVIDIA hardware.
Developers can also access the model through hosted providers or download its weights from model catalogs. Those alternatives let teams evaluate or deploy it without adopting the Mac Studio configuration.
Downloadable weights come with license conditions
Meta is distributing the weights under the Muse Model License, which it says includes limited commercial terms allowing many developers to use the model for free. Open-weight availability should therefore not be read as unrestricted permission for every commercial application.
The launch establishes technical availability, not a forecast for Apple hardware sales or Meta revenue. Neither announcement provides a quantified financial contribution from the release.
Conclusion
Muse Spark Open adds a local deployment option for developers with suitable hardware, alongside hosted access. Meta says a full evaluation report will follow, but it has not specified a publication date.
Read More News
Disclaimer
This FinanceVelix article provides general information, not personalized financial, investment, legal, or tax advice. Market prices, interest rates and economic conditions can change; investments carry a risk of loss. Facts reflect the material available when published and may change. Check important information independently and consult a qualified professional when needed. Mentioning an asset, company, or service is not an endorsement.
