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
Muse Spark Open puts an open-weight multimodal model on your desk: 816B total / 42B active parameters, 1M-token context, and local inference on two M5 Ultra Mac Studios.
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
Why Run Muse Spark Open
Released under open weights, Muse Spark Open is a natively multimodal system built for coding, autonomous agents and long-horizon jobs — and you can host it on your own hardware.
- Weights You Can OwnBuilt in the United States and shipped as the most capable open-weight release so far, Muse Spark Open pairs 816B total parameters and 42B active ones with a million-token window and native multimodality.
- Frontier Power at Your DeskNo datacenter lock-in required: Muse Spark Open starts up on a workstation of two Apple M5 Ultra Mac Studios running the latest MLX build, keeping frontier-class inference under your own roof.
- Agentic Results, Proven LocallyMuse Code on top of Muse Spark Open delivered working iOS fixes and features to Ice Cubes, then operated DaVinci Resolve from start to finish and reframed the best shots into vertical crops.
How to Run Muse Spark Open
Choose your path with Muse Spark Open: local Apple silicon, an inference engine that suits your stack, or a hosted endpoint.
What Muse Spark Open Can Do
From a million-token window and built-in multimodality to desktop-grade memory bandwidth and open weights, these are the traits that let Muse Spark Open handle agentic video work.
A Million Tokens of Context
With a 1M-token window, Muse Spark Open keeps long agentic sessions and multi-step reasoning chains connected to everything that came before.
Agentic Coding That Runs Long
Tuned for software engineering and autonomous loops, Muse Spark Open takes on long-horizon jobs and drives tools like DaVinci Resolve from beginning to end.
Multimodality Baked In
Rather than relying on add-on adapters, Muse Spark Open interprets mixed inputs natively — which is exactly what enables frame-level review of footage.
Two Mac Studios, One Supercomputer
Unified memory gives each M5 Ultra GPU access to as much as 512 GB, so two Mac Studios host the entire Muse Spark Open model as a single machine with a terabyte of pooled memory.
1.2 TB/s of Memory Bandwidth
The M5 Ultra sustains 1.2 TB/s of bandwidth — enough to feed the small per-token slice and keep Muse Spark Open responsive on a desktop.
Safety Post-Training Built In
Open weights come with extra safety post-training: Muse Spark Open was measured against the Meta Superintelligence Scaling Framework and met its release threshold across the risk categories assessed.
Muse Spark Open — FAQ
Answers to frequent questions on downloading, licensing and running Muse Spark Open.
What hardware do I need to run Muse Spark Open locally?
Two Apple M5 Ultra Mac Studios running the latest MLX build — each one supports up to 512 GB of unified memory, and combined they provide roughly a terabyte for the full model.
How large is the model and how much context does it handle?
Muse Spark Open ships with 816B total parameters and 42B active, a 1M-token context window and native multimodality, targeting coding, agentic work and long-horizon tasks.
Where can I download the weights or access the model via API?
You can download weights from Amazon SageMaker AI, Google Cloud Model Garden, Hugging Face and Microsoft Foundry Models, while hosted APIs are available at Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter and Together AI.
Which inference engines and chips are supported?
Choose from MLX, llama.cpp, Unsloth, SGLang, vLLM, TensorRT-LLM or Transformers — optimized for NVIDIA, AMD, Intel and Apple hardware, with PyTorch underneath whenever you want to dig deeper.
What license applies to the released weights?
The Muse Model License governs the release; its limited commercial terms still allow a wide range of developers to run Muse Spark Open at no cost.
How was the model evaluated for safety before release?
It was assessed against the Meta Superintelligence Scaling Framework for the most severe risks, received additional safety post-training for open weights, and passed the release thresholds — full details are in the model card.
Start Building with Muse Spark Open Weights
Pull Muse Spark Open from Hugging Face, query it through hosted inference partners, and tell the community what you create on the Hugging Face Forums.