Hacker News new | past | comments | ask | show | jobs | submit
skohan | on: Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows
Yes they have quants for 32GB and 20GB use-cases (including mmproj and kv cache + context)