CPU backends precisions support

[truecoder34]

Hi all,

Does CPU backends supported in ggml (ZDNN, ZenDNN, BLAS, KleidiAI) only able to run inference of models with F16/B16 and F32 precision only, and any quantized models will be run with Naive CPU backend?
Is this limitation actual for all CPU Backends excluding Naive CPU one ?

Thank you in advance .