None
None
Long Story Video Skill
Turn your idea into 10s–10min video—AI writes scripts, prompts, and generates footage automatically.
Ads Video Skill
Generate professional ads and sales videos—AI auto-generates scripts, prompts, and footage.
3D Science Explainer Video Skill
Convert scientific concepts into stunning 3D explain animations
Feedback
freeTrialImage.bannerPity
freeTrialImage.upgradeUnlock
- ✓freeTrialImage.benefitHd
- ✓freeTrialImage.benefitWatermark
- ✓freeTrialImage.benefitUnlimited
Muse Spark Open
Download the open Muse Spark Open weights and run multimodal, million-token agentic video edits on your own Apple silicon workstation.
All Tools
Discover our comprehensive AI-powered animation toolkit
Muse Video
Seedance2.5
The Future of AI Video Is Here.
Seedance 2.0
The Future of AI Video Is Here.
Veo3.1
Create Stunning Videos with Veo3.1
Grok Video Generator
Create Videos from Text or Images with AI
MiniMax H3 video generator
MiniMax H3 AI Video Generator
Seedance 2.0
The Future of AI Video Is Here.
Gemini Omni 1.2 Flash
Muse Spark Open, Running on Hardware You Control
The weights are public: Muse Spark Open is a natively multimodal system built for coding, agentic pipelines and long-horizon jobs that you can serve on your own machines.
- Fully Open Weights, Made in the USThis release ships as the most capable open-weight model to date from a US lab: 816B total parameters, 42B active, a 1M-token window and multimodality built in natively.
- Supercomputer Power at Your DeskFrontier models no longer have to live behind a datacenter door — Muse Spark Open boots on a pair of Apple M5 Ultra Mac Studios running the latest MLX build.
- Agentic Tasks Verified On-DeviceRunning Muse Code, it pushed real iOS fixes and features into Ice Cubes, then operated DaVinci Resolve from start to finish and reframed the best shots vertically.
Three Ways to Deploy Muse Spark Open
Choose local Apple silicon, an inference engine you prefer, or a hosted API partner — each path gets Muse Spark Open running in minutes.
What Muse Spark Open Can Do
A million-token window, multimodality built in, desktop-class memory bandwidth and public weights — these are the pieces that let Muse Spark Open handle agentic video work.
A One-Million-Token Context
With a 1M-token window, Muse Spark Open keeps earlier steps in view across long agentic sessions and multi-step reasoning chains.
Agentic Coding Over Long Horizons
Tuned for software engineering and agent loops, Muse Spark Open sustains long-horizon jobs and drives tools such as DaVinci Resolve from end to end.
Multimodality Built In
Muse Spark Open ingests mixed inputs natively instead of relying on bolt-on adapters — the very reason frame-by-frame footage review works at all.
A Pair of Mac Studios as One Machine
Unified memory lets each M5 Ultra GPU address up to 512 GB, so two Mac Studios hold the entire Muse Spark Open model with a terabyte of pooled memory.
1.2 TB/s of Memory Bandwidth
The M5 Ultra pushes 1.2 TB/s — enough to feed the small per-token slice and keep Muse Spark Open responsive on a desktop setup.
Extra Safety Post-Training
Open weights receive additional safety post-training: Muse Spark Open was measured against the Meta Superintelligence Scaling Framework and passed its release bar across the risk categories tested.
Muse Spark Open FAQ
Answers to the questions people ask most about downloading, licensing and running Muse Spark Open.
What hardware does Muse Spark Open need to run locally?
Two Apple M5 Ultra Mac Studios with the latest MLX build — each offers up to 512 GB of unified memory, giving roughly a terabyte together for the full model.
How big is the model and how much context can it hold?
Muse Spark Open has 816B total parameters with 42B active, a 1M-token context window and multimodality built in, designed for coding, agentic work and long-horizon tasks.
Where can I get the weights or call the model through an API?
Weights are available on Amazon SageMaker AI, Google Cloud Model Garden, Hugging Face and Microsoft Foundry Models; hosted APIs are offered by Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter and Together AI.
Which inference engines and chips does it support?
Choose from MLX, llama.cpp, Unsloth, SGLang, vLLM, TensorRT-LLM or Transformers — tuned for NVIDIA, AMD, Intel and Apple hardware, with PyTorch underneath when you want to go deeper.
What license covers the released weights?
The Muse Model License applies; its limited commercial terms still let a broad range of developers run Muse Spark Open at no cost.
How was safety evaluated before the release?
It was scored against the Meta Superintelligence Scaling Framework for the most severe risks, received extra safety post-training for open weights, and cleared the release thresholds — the model card has the details.
Start Building With the Muse Spark Open Weights
Pull Muse Spark Open from Hugging Face, call it through hosted inference partners, and tell the community what you built on the Hugging Face Forums.