Meta Open-Sources Muse Glimmer: A 30B Local Agent Model with 3.1x Speed

dl_weekly · x · 2026-08-19

Meta has released Muse Glimmer, a 30-billion-parameter open-source model optimized for local agentic workflows. Quantized to under 20GB, it runs on consumer GPUs and achieves up to 3.1x faster decoding via speculative decoding. It is licensed under Apache 2.0 and supports integrations with llama.cpp and MLX.

Related event: Meta Open-Sources Muse Glimmer, a 30B Local Voice Agent Model(2 posts)→

Original post →

More from Infra

Infra channel →