Post #2749991

#ArchLinux managed to release ollama-cuda 0.23.0 missing the important (and only file the package provides) /usr/lib/ollama/libggml-cuda.so so currently the latest #ollama do not have support for #cuda and your #LLM will be far slower. Issue may be the unsupported #GCC version 16 (think the official supported gcc version was 13).