Muse Spark Open
The most capable US open weights model to date
36 followers
The most capable US open weights model to date
36 followers
Meta's open weights model built for coding and agentic work. 816B total parameters with 42B active, a 1M token context window and native multimodal input. Download the weights on Hugging Face or run it on two M5 Ultra Mac Studios with MLX.
Launch Team
CloudflareYour next app belongs on Cloudflare Workers. Deploy for free
Promoted
Hey PH community 👋
Excited to hunt Muse Spark Open today.
Meta just released the weights for Muse Spark Open, which they call the most capable US open weights model to date. It has 816B total parameters with 42B active, a 1M token context window and native multimodal input, and Meta says it was built for coding and agentic work like long horizon tasks and complex software engineering. Mark posted a look at it running on two Mac Studios here.
Running on your own desk is the part I'm most excited about. Meta worked with Apple so the full model fits across two M5 Ultra Mac Studios using MLX, and they tested real agent work fully on device, from shipping iOS fixes to the open source Mastodon app Ice Cubes to editing footage in DaVinci Resolve.
Agentic work is where Meta puts the headline numbers. They report 89.2 on MCP Atlas and 76.4 on Terminal Bench 2.1, and the model card frames it as a foundation for coding agents that plan, call tools and recover from failures over long workflows.
For builders, it's already live through Together AI, Fireworks, Groq, Baseten, Modal, Nebius and OpenRouter, and it runs on SGLang, vLLM, llama.cpp and Unsloth under the Muse Model License.
You can grab the weights on Hugging Face right now: https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B
Give it a spin and let the team know what you create with Muse Spark Open 🚀
PS: Tagging @alexandrw and @dps_ if you're here for any additional comments/ thoughts.