Saw someone here the other day saying normal llama supports all the iq quant variants now and its faster than ik_llama too.
Well i just went to the trouble of updating and recompiling my copy and no it does not, fuck you faggot