Good question! I am collecting human data on how quantization affects outputs. See here for more information: ggml-org/llama.cpp#5962

In the meantime, use the largest that fully fits in your GPU. If you can comfortably fit Q4_K_S, try using a model with more parameters.

See the wiki upstream: https://github.com/ggerganov/llama.cpp/wiki/Feature-matrix