Meta is back with Muse Glimmer: local, agentic, multimodal, and open source
Muse Glimmer 30B is a multimodal model with 30 billion parameters, capable of handling tasks like object detection, video analysis, and tool calling. It supports efficient inference using llama.cpp, DFlash speculative decoding, and OpenAI-compatible APIs, making it suitable for low-latency, cost-effective deployment. The model offers robust capabilities in image and video understanding, with features like video question answering and end-to-end object detection.
More from Huggingface
LFM2.5-VL-3B for Better and Faster Vision Capabilities for the Edge
Thinking of ACE? We Can Do It with Fewer Tokens
Making Knowledge Distillation Cheap Enough to Run at Scale
