Meta Open-Sources 30B AI Model Muse Glimmer for Local Real-Time Agents

Meta released and open-sourced its AI model Muse Glimmer on August 10. It has 30 billion parameters and is available under the Apache 2.0 license. The model is designed for always-on local agent workflows. Using 4-bit quantization, it can run on Mac or PC with 24GB or 32GB of VRAM, with almost no performance loss.

Tests on MacBook M4 Max, M5 Max, and RTX 5090 showed smooth conversations and real-time agent interactions, all processed locally. Compared to Gemma4-31B and Qwen3.6-27B, Muse Glimmer showed strong performance in several benchmarks. The weights are available on Hugging Face.