# 4-bit quantization compression

Molmo-7B-D is a large language model quantized with BnB 4-bit. The model size is reduced from 30GB to 7GB, and the video memory requirement is reduced to about 12GB.