thank you for your contribution, may I ask for Qwen3.5-27B oQ4?
#1
by rairenter - opened
Hello
I love qwen3.5 and omlx a lot for speed & quality, 35B-A3b and 27B are my favourite model that use locally. Can you also convert 27b to oQ4?
anyway, thank you for your model
Hi.
I love it too! Here you go: Qwen3.5-27B-MLX-oQ4
I have also reconverted this Qwen3.5-35B-A3B-MLX-oQ4 with the latest oMLX v0.2.24 just in case... and had to reupload it because all the artifacts got changed. Things are moving really fast.
Updated the readme file with "Sensitivity model" I used + there's no "Group Size" setting anymore (it was confusing due to oQ dynamic nature, I suppose).
deepsweet changed discussion status to closed