muse-glimmer

Muse Glimmer 30B

Muse Glimmer is Meta's 30B parameter open model built for local, agentic AI. It combines multi-step reasoning, reliable tool use, multimodal understanding, and failure recovery in a model designed to run on consumer hardware without requiring cloud infrastructure.

Highlights

• Optimized for local agents: designed for autonomous workflows on high-end PCs and workstations, including multi-step planning, sequential tool invocation, and long-horizon task completion.
• Strong agentic capabilities: trained and evaluated for tool use, coding agents, scaffold compatibility, and recovery from failed or unexpected tool results.
• Multimodal input: uses a dedicated perception encoder to understand interleaved text and images, including screenshots, charts, and documents.
• Controllable reasoning: supports configurable reasoning strength so users can balance quality and speed across simple and complex tasks.
• Long-context and permissive: supports a 131K context window and is released under the Apache 2.0 license.