RTX Pro 6000, ends up taking ~66GB when running the MXFP4 native quant with llama-server/llama.cpp and max context, as an example. Guess you could do it with tw

RTX Pro 6000, ends up taking ~66GB when running the MXFP4 native quant with llama-server/llama.cpp and max context, as an example. Guess you could do it with two 5090s with slightly less context, or different software aimed at memory usage efficiency.