Meta Open-Sources Muse Glimmer: 30B Agentic Model That Fits a 24GB Consumer GPU
bibryam · x · 2026-09-26
Meta Superintelligence Labs released Muse Glimmer, a 30-billion-parameter model open-sourced under Apache 2.0, designed for always-on local agent workflows.
- Deployment: its 17GB quantization fits a single 24GB consumer GPU with 1% average degradation across 15 benchmarks; optimized llama.cpp, MLX and ExecuTorch integrations are landing soon.
- Focus: local agents, function calling, local coding and LLM-as-a-judge evaluation — fully offline, no cloud required.
- Weights are on Hugging Face with developer documentation; the blog details how capabilities like long-horizon task execution were trained together.
It's Meta's latest bet that well-trained small models can approach frontier performance on targeted agentic tasks.
More from Infra
- Running Qwen2.5-Coder-3B on Android: Jupyter + Pydantic Evals on-device eval pipeline — alex_b3b · 2026-11-11
- 22% of Nvidia revenue now books to Singapore, but only ~1% is local buyers — Gregory_C_Allen · 2026-11-11
- Huawei's Kirin 9050 NPU claims offline 30B MoE on smartphones, +140% perf — AIFlow_ML · 2026-11-11
- Stratechery: Intel vs AMD's diverging earnings and Intel's need for an old-new path — Stratechery · 2026-11-11
- a16z Podcast: Mike Volpi on What the 1990s Internet Boom Teaches Us About the AI Buildout — a16z Podcast · 2026-11-11
- Alibaba files to list chip-design unit T-Head in Hong Kong to fund AI push — pstAsiatech · 2026-11-11