Bigger model wins as long as the quantization was done properly.