Running Meta's Muse Glimmer 30B Locally on Mac Studio at 30 tok/s

MaziyarPanahi · x · 2026-08-10

A developer has successfully run Meta's newly released Muse Glimmer 30B multimodal model (GGUF format) locally on a Mac Studio.

A recorded demo shows a real two-turn conversation where the model streams its reasoning live before landing on the answer, measuring about 30 tokens/second via the local API. The author is currently testing the local stack across healthcare workflows, visual pipelines, tool use, and full agentic tasks.

Related event: Meta's Muse Glimmer 30B Runs Locally on Mac Studio with Day-Zero Support(2 posts)→

Original post →

More from Models

Models channel →