AI & Enterprise
Naver team builds 4,000 Nvidia Blackwell B200 GPU cluster, says AI model development 12 times faster
Naver said on Jan. 8 it has completed an AI computing cluster based on 4,000 Nvidia B200 Blackwell GPUs. The company said the new “B200 4K cluster” is comparable in computing scale to leading TOP500 supercomputers. Internal simulations showed training a 72 billion-parameter model could be cut from about 18 months on an A100-based 2,048-GPU setup to about 1.5 months, improving training efficiency by more than 12 times.