tested the quantized models, both load correctly as of 41aee4d in llama.cpp.