Meta Releases Muse Glimmer, a 30B Open-Weights Local Agentic Model
Meta Superintelligence Labs has released Muse Glimmer, a 30-billion parameter open-weights multimodal model designed specifically for local agentic workflows and coding. Released under the Apache 2.0 license, the model supports interleaved text and image inputs, is trained on over 100 languages, and features controllable reasoning effort. This release strengthens the open-weights AI ecosystem by providing a powerful, agent-focused model that can run locally, offering developers an alternative to proprietary frontier models. It highlights the growing trend of distilling larger frontier models into smaller, highly capable local models for agentic tasks. Muse Glimmer is a dense model that requires over 55GB of memory at full precision, making local execution on consumer hardware challenging without quantization. It is a distilled version of Meta's larger Muse Spark model and is already available on platforms like Ollama.
## BACKGROUND
"Open-weights" models allow developers to download and run the model's parameters locally, offering greater privacy and customization compared to API-only models. "Agentic AI" refers to systems that can autonomously execute complex, multi-step tasks and use tools with minimal human supervision, which typically requires strong reasoning and coding capabilities.