Start building

Meta's Muse Spark open-weight release Is Now in Limited Preview on GroqCloud

Meta's Muse Spark Open Now Available in Limited Preview on GroqCloud

We're excited to bring Muse Spark Open, Meta's new open-weight model for agentic coding, to GroqCloud on launch day. It is available today in limited preview.

Key Highlights

  • 816B total parameters with 42B active per token, in a Mixture of Experts design with 256 experts
  • 131,072-token context window supported on GroqCloud, enough to work with substantial codebases, including dependencies and legacy code
  • Text and image input, with text output, on GroqCloud
  • Trained on real developer workflows: multi-file refactors, debugging sessions, and long-running tasks
  • 350+ tokens per second per user on GroqCloud

Built for Agentic Coding

Meta built Muse Spark Open as an agentic model from the start. It plans, acts, observes, and corrects course over long tasks. It can spawn and coordinate subagents, browse the web, and use computers across desktop, mobile, and browser. Before each action it writes a short statement of intent, so developers can follow what it is doing and why. Meta reports higher first-attempt accuracy and reliable tool calling.

What This Means for Developers

An agentic coding task is a chain of model calls, and each step waits on the one before it. Inference speed therefore decides how long a developer waits for the finished result. On GroqCloud, Muse Spark Open runs at 350+ tokens per second per user.

More than six million developers build on GroqCloud, along with Fortune 500 enterprises and thousands of AI-native companies. The platform runs trillions of tokens every week from data centers in North America, Europe, the Middle East, and APAC.

"Meta built Muse Spark Open for the long, multi-step coding work that developers are handing to agents today," said Scott Albin, VP, Head of Commercial & Customer at Groq. "Those agents make dozens of model calls to finish a single task. Running the model on GroqCloud keeps each of those calls fast, so developers get work done faster."

Interested in getting access? Head over to https://groq.com/muse-spark-open

Media Contact

pr-media@groq.com