Hacker News new | past | comments | ask | show | jobs | submit
eugene3306 | on: Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows
will it run on 2x 5060Ti with 16GB each?
skohan | parent | next
It should - the kquant-dynamic variant is targeted towards 32GB. Downloading it now to give it a try.
leansensei | parent | next
It does, beautifully. Now let's wait for an NVFP4 GGUF!
BoredomIsFun | parent
yes. you can even parallelize two cards and get 1.7 times the speed.