Hacker News new | past | comments | ask | show | jobs | submit
folienumero | on: Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows
In my experience it's faster (10tk/s vs 35tk/s) and better than qwen3.6 series.