[hedgehog0]

May I ask how small? From what I see on r/localllama, it seems that at least 3B and 7B ones are starting to be useful? Also I cannot use CUDA or modern AMD cards..