Meta Superintelligence Labs has released Muse Spark Open, the open-weight version of its Muse Spark model.
Key points
- Meta released Muse Spark Open, an 816B-parameter MoE model.
- The model has a 1 million token context window.
- It accepts multimodal inputs: text, image, video, audio.
- Muse Spark Open runs on two Apple M5 Ultra Mac Studios.
- It shows strong performance in agentic tool use.
This new model is an 816 billion parameter Mixture-of-Experts (MoE) model that activates 42 billion parameters per token. It is designed to process text, image, video, and audio inputs, returning text outputs, and boasts a 1 million token context window.
Meta claims Muse Spark Open is the most capable US open-weight model released to date. For developers, a key feature is its deployment capability, with the NVFP4 build running on two Apple M5 Ultra Mac Studios. The model's architecture includes 62 layers and a hidden dimension of 8192, with each MoE layer containing 256 experts and activating 8 per token through top-k routing.
By the numbers
- Total parameters
- 816B
- Active parameters per token
- 42B
- Token context window
- 1M
The model supports multimodal input through a separate 1.8 billion parameter Perception Encoder. Its vocabulary size is 202,048 tokens, and it was trained on data covering over 100 languages. Muse Spark Open offers adjustable reasoning depth settings, defaulting to 'xhigh'.
Performance benchmarks show Muse Spark Open excelling in agentic tool use and knowledge reliability, achieving a score of 89.2 on MCP-Atlas. However, it performed mid-pack on coding tasks, scoring 76.4 on Terminal-Bench, which is behind models like Kimi K3 (85.0), GLM 5.3 (83.9), and Qwen3.8 (82.0).
Sources and references
Original report
Meta Releases Muse Spark Open: 816B Open-Weight MoE With 1M Context
MarkTechPost
Read on MarkTechPost (opens in a new tab)Linked in the original report
- Linkresearch.meta.ai/blog/muse-spark-open (opens in a new tab)
- Postlinkedin.com/posts/today-were-releasing-muse-spark-o… (opens in a new tab)
- Modelhuggingface.co/meta-models/Muse-Spark-Open-816B-A42B (opens in a new tab)
- Modelhuggingface.co/meta-models/Muse-Spark-Open-816B-A42B-N… (opens in a new tab)
- Linkresearch.meta.ai/static/muse-spark-open-evaluation-metho… (opens in a new tab)
- Linkdev.meta.ai/resources/blog/build-with-muse-spark-open (opens in a new tab)
Ref 00L2PM5P
Cite this
APA
Interestana AI Editorial. (2026, October 16). Meta Releases Muse Spark Open AI Model. Interestana. https://interestana.com/articles/meta-releases-muse-spark-open-816b-open-weight-moe-with-1m-context-00l2pm5p
MLA
“Meta Releases Muse Spark Open AI Model.” Interestana, 16 Oct. 2026, interestana.com/articles/meta-releases-muse-spark-open-816b-open-weight-moe-with-1m-context-00l2pm5p.
Chicago
Interestana AI Editorial. “Meta Releases Muse Spark Open AI Model.” Interestana, October 16, 2026. https://interestana.com/articles/meta-releases-muse-spark-open-816b-open-weight-moe-with-1m-context-00l2pm5p.
BibTeX
@misc{interestana_00l2pm5p,
author = {{Interestana AI Editorial}},
title = {{Meta Releases Muse Spark Open AI Model}},
howpublished = {Interestana},
year = {2026},
month = oct,
url = {https://interestana.com/articles/meta-releases-muse-spark-open-816b-open-weight-moe-with-1m-context-00l2pm5p},
note = {Summary of reporting by MarkTechPost, https://www.marktechpost.com/2026/10/15/meta-muse-spark-open-816b-open-weight-model/}
}Credit the source
Original reporting: MarkTechPost, “Meta Releases Muse Spark Open: 816B Open-Weight MoE With 1M Context”, https://www.marktechpost.com/2026/10/15/meta-muse-spark-open-816b-open-weight-model/(opens in a new tab)
Follow this story
Get Artificial Intelligence in your inbox — new stories in 40 words or fewer, once a week.
Subscribe