Journalism begins where hype ends

,,

The use of AI is never a purely technical matter: when it enters processes that affect people’s lives, it touches on rights, opportunities, status and freedom"

—Pope Leo XIV

Meta Releases Muse Spark Open, an 816B-Parameter Open-Weight AI Model

Meta’s Muse Spark Open is described as an 816-billion-parameter open-weight AI model with native multimodal capabilities, a 1-million-token context window and support for coding, agentic workflows and complex reasoning.
Meta Muse Spark Open AI model announcement graphic featuring the Meta logo and blue flowing lines.
October 15, 2026 11:15 PM IST | Written by Vaibhav Jha

Meta announced the release of Muse Spark Open, an open-weight natively multimodal AI model with 816B total parameters (42B active) and a 1M-token context window, natively multimodal across text, image, video, audio and optimized to run on Mac Studio with M5 Ultra.

Meta released the weights on Hugging Face and Amazon SageMaker AI, Google Cloud Model Garden, and Microsoft Foundry Models under the Muse Model License with limited commercial terms.

Meta’s AI team has positioned Muse Spark Open as their most-capable US open-weight AI model yet, engineered for coding, agentic workflows, long-horizon tasks, complex software engineering, and multi-step reasoning.

 

Marking a notable return toward open weights after Meta’s proprietary Muse Spark series launched earlier in 2026, Muse Spark Open is optimized to run on two Apple M5 Ultra Mac Studios using Apple’s MLX framework, leveraging unified memory (up to 512 GB per machine).

The unique feature enables the pair to function like a single supercomputer with roughly a terabyte of shared memory and high bandwidth, claimed Meta.

“Muse Spark Open is available today through API platforms at Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter and Together AI. Model weights are available for download on Amazon SageMaker AI, Google Cloud Model Garden, Hugging Face, and Microsoft Foundry Models. It runs on leading inference engines like SGLang, vLLM, TensorRT-LLM, Hugging Face Transformers, Apple MLX, llama.cpp, and Unsloth with optimizations for AMD, Apple, Intel, and NVIDIA silicon. Developers can easily further customize the model using the PyTorch framework,” read a statement from Meta.

The release of Muse Spark Open AI model comes in the wake of Meta’s ambitious chase of “superintelligence” and its advocacy for open-weight AI models. Recently, it had released Muse Glimmer, a 30B open-weight AI model.

Also Read: Meta Launches Muse Wonder and Muse Wonder Full Frontier AI Models

Author

  • Vaibhav Jha, editor and co-founder at AI FrontPage

    Vaibhav Jha is an Editor and Co-founder of AI FrontPage. In his decade long career in journalism, Vaibhav has reported for publications including The Indian Express, Hindustan Times, and The New York Times, covering the intersection of technology, policy, and society. Outside work, he’s usually trying to persuade people to watch Anurag Kashyap films.

    LinkedIn