gemma-2

Google launches Gemma 2, its open-weight model family, in 9B and 27B sizes

The 27B model ran on a single H100 GPU or TPU host, which Google said cut deployment cost while matching models more than twice its size.

Open weights & ecosystem