Editorial illustration for Meta releases Muse Spark Open, an open-weight AI model designed for Macs
Meta Releases Muse Spark Open for Mac AI
Meta put out a new open-weight model Thursday called Muse Spark Open, and the headline number is 816 billion parameters paired with a 1 million token context window. The company is calling it the most capable open-weight model built in the United States, and it's specifically tuned to run on Apple hardware, a departure from the usual server-farm assumptions behind most large language models.
That local-first design comes with a catch. Engadget reports the model needs at least 1 terabyte of RAM to run, which means most Mac owners are out of luck. Meta says it performs best on a Mac Studio with the new M5 Ultra chip, hardware that can actually supply that kind of memory. Muse Spark Open is natively multimodal, built for coding work and agentic tasks, and Meta paired it with Muse Code, its AI coding terminal, to test it against real jobs like patching a Mastodon client and running edits inside DaVinci Resolve.
The release also reopens the usual argument about why companies bother shipping open-weight models at all when closed systems dominate headlines.
Muse Spark Open is natively multimodal and built for complex coding tasks and agentic labor. It has also been designed to run locally on Apple’s computers. However, you will still need a lot of power to run the model locally, with a requirement of at least 1 terabyte of RAM, per Engadget.
Why this matters
A 816-billion-parameter model with a 1 million token context window optimized for Apple's hardware is a notable shift, not just a spec bump. Developers building on Macs have mostly been stuck running smaller quantized models or routing calls to cloud APIs; if Muse Spark Open genuinely performs well on Apple silicon, that changes local development workflows for a lot of teams, especially ones with budget constraints or privacy requirements that rule out sending data to third-party servers. The "most capable open-weight model in the United States" framing is Meta's own, so we'd want to see independent benchmarks before taking that at face value, particularly against Llama's own prior releases and other open-weight contenders.
Still, the multimodal claims and the coding focus suggest Meta is chasing the same developer audience that's been drifting toward Chinese open-weight labs. Worth watching: how it actually runs on consumer Mac hardware versus Meta's reference setups, and whether "open-weight" here comes with licensing terms as permissive as the marketing implies.
Related Reading
Browse all Meta news →
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
Related Topic Meta’s Muse Tracks Your Email and Bank Data While You Work
AI Tools & Apps
Related Topic Meta Faces Lawsuit Over AI Training Data, Facial Recognition
Policy & Regulation
From Archives Bluesky expands AI app Attie to let anyone build websites, games and more
From Archives Rivian’s AI agent to gain access to frontier AI products, CEO says
From Archives Meta Muse Spark Open on NVIDIA GB300 NVL72 Scales Agentic Coding Workflows
AI Tools & Apps
From Archives Meta Muse Wonder vs Claude: Rival Runs Best-Effort, Not Common Harness
Common Questions Answered
What are the key specifications of Meta's Muse Spark Open model?
Muse Spark Open features 816 billion parameters paired with a 1 million token context window, making it Meta's most capable open-weight model built in the United States. The model is natively multimodal and specifically optimized to run locally on Apple hardware, representing a significant departure from traditional server-based large language models.
What are the hardware requirements needed to run Muse Spark Open locally on a Mac?
Running Muse Spark Open locally requires at least 1 terabyte of RAM on your Apple computer. This substantial memory requirement reflects the model's 816 billion parameters and advanced capabilities for complex tasks.
How does Muse Spark Open change the development workflow for Mac-based developers?
Previously, Mac developers were limited to running smaller quantized models or routing API calls to cloud services, which created latency and privacy concerns. With Muse Spark Open optimized for Apple silicon, developers can now run a highly capable model locally, enabling better workflows for teams with budget constraints or strict privacy requirements that prevent sending data to third-party servers.
What types of tasks is Muse Spark Open designed to handle?
Muse Spark Open is built specifically for complex coding tasks and agentic labor, combining its multimodal capabilities with the processing power needed for sophisticated development work. Its 1 million token context window allows it to handle extensive code repositories and complex problem-solving scenarios.
Further Reading
- Meta launches an open-weight model that runs on Macs - TechCrunch
- Meta built a huge open-weight AI model for Macs, but you'll need 1TB of RAM to run it - Engadget
- Apple and Meta team up to make Muse AI model run on Mac computers - CNBC
- Meta and Apple Bring Muse Spark Open AI Model to Mac Studio - MacRumors
- Welcome Muse Spark Open: open-weight frontier multimodal mixture-of-experts model from Meta - Hugging Face
