[terafo]

It fits, whisper.cpp uses 4 bit quantization, 13B model takes a little bit more than 8gb and around 9gb ram while inferencing.