@karebu i'm using my 6700XT for it and it's performing like any other 14b model. the difference is that it will display the thought process as regular text so for an actual response you need to wait a bit longer
@karebu if you're hosting ollama in docker and your amd gpu isn't officially supported (anything below 6800 i think?) then you pass an environment based on which GPU you've got (for me it's HSA_OVERRIDE_GFX_VERSION=10.3.0, you can find them here)
