← AI Glossary

GPU

Infrastructure & deployment

A GPU, or graphics processing unit, is a chip designed to run thousands of calculations in parallel. It has become essential for training and running AI models. Its memory, the VRAM, determines how large a model it can load: a 70-billion-parameter model requires a dedicated GPU server.

In practice at Gensai

Thanks to quantization, Gensai runs some models without a GPU server, like the “Monsters of the Oceans” installation on a simple Mac Mini.