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
Explore Muse Spark Open — openly licensed weights, 816B parameters, a million-token window, and local inference on Apple silicon.
All Tools
Discover our comprehensive AI-powered animation toolkit
MiniMax H3
MiniMax H3 AI Video Generator
Seedance 2.5
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
FLUX 3 Video Generator
Kling 3.0
Next-Gen AI Video Generator
Grok Video Generator
Create Videos from Text or Images with AI
MiniMax H3 video generator
MiniMax H3 AI Video Generator
Inside the Muse Spark Open Release
Shipped as downloadable weights, this natively multimodal system takes on coding, autonomous workflows and long-horizon jobs on hardware you control yourself.
- Downloadable Weights, US-DevelopedThis launch delivers the strongest openly available model built in the United States — 816B total parameters, 42B active, a million-token window and native multimodality.
- Frontier Power at Your DeskTop-tier models should not stay walled inside a datacenter; this one boots on your own workstation when paired with a couple of Apple M5 Ultra Mac Studios and the newest MLX build.
- Agentic Tasks, Verified LocallyMuse Code running on this model delivered genuine iOS fixes and features to Ice Cubes, then operated DaVinci Resolve from start to finish, reframing the best shots vertically.
Ways to Deploy Muse Spark Open
Three paths to get the model working — local Apple silicon, the inference engine you prefer, or a managed hosted endpoint.
Capabilities Inside Muse Spark Open
A million-token window, native multimodality, desktop-grade memory bandwidth and downloadable weights — the ingredients behind long-running agentic video work.
Million-Token Context
A 1M-token window keeps lengthy agentic sessions and multi-step reasoning chains anchored to earlier context instead of losing the thread.
Agentic Coding Over Long Horizons
Built for software engineering and agentic loops, Muse Spark Open sustains long-running jobs and drives tools such as DaVinci Resolve from start to finish.
Multimodal by Design
Mixed inputs are read natively rather than through add-on adapters, which is exactly what enables frame-by-frame footage review.
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 full Muse Spark Open model with a terabyte of pooled memory.
1.2 TB/s of Memory Bandwidth
The M5 Ultra shifts 1.2 TB/s — plenty to feed the small per-token slice and keep responses fast on a desktop.
Safety Tuned Before Release
Open releases get extra safety post-training: the model was evaluated against the Meta Superintelligence Scaling Framework and met its release bar across measured risk categories.
Muse Spark Open: Common Questions
Straight answers on running, licensing and downloading the open weights.
Which hardware is required for local inference?
Two Apple M5 Ultra Mac Studios running the latest MLX build — each addresses up to 512 GB of unified memory, giving roughly a terabyte together for the full Muse Spark Open model.
What are the model size and context limits?
It ships with 816B total parameters (42B active), a 1M-token context window and native multimodality, targeting coding, agentic workflows and long-horizon tasks.
Where do I get the checkpoints or an API endpoint?
Weights are downloadable via Amazon SageMaker AI, Google Cloud Model Garden, Hugging Face and Microsoft Foundry Models; managed APIs run at Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter and Together AI.
What runtimes and hardware are supported?
Choose MLX, llama.cpp, Unsloth, SGLang, vLLM, TensorRT-LLM or Transformers — tuned for NVIDIA, AMD, Intel and Apple silicon, with PyTorch underneath when you want to go deeper.
Which license covers the released weights?
The Muse Model License applies; its limited commercial terms still leave plenty of room for a wide range of developers to run Muse Spark Open at no cost.
How was safety evaluated prior to launch?
It was scored against the Meta Superintelligence Scaling Framework covering the most severe risks, received additional safety post-training for open weights, and passed the release thresholds — the model card has the details.
Download the Muse Spark Open Checkpoints
Pull Muse Spark Open weights from Hugging Face, query it through hosted inference partners, and share what you build on the Hugging Face Forums.