Gemma 2 at a glance
Google said Gemma 2 27B competed with models more than twice its size and ran at full precision on a single H100 GPU, under a license that allowed commercial use.
What changed with Gemma 2
Google released Gemma 2 on June 27, 2024, in 9B and 27B parameter sizes, with a 2.6B model announced for later. Google said the 27B model competed with models more than twice its size and that the 9B model outperformed Llama 3 8B and other open models in its class.
The 27B model could run inference at full precision on a single NVIDIA H100 GPU or TPU host. Weights were available on Kaggle and Hugging Face under the Gemma license, which permits commercial use, and the 27B model could be tried in Google AI Studio.
Where to compare Gemma 2
Independent leaderboards and official resources for comparing this model yourself. Rankings change often; each site shows its own current data.
General leaderboards
Why Gemma 2 matters for businesses
For businesses, Gemma 2 lowered the hardware bar for private deployments: a strong model on one GPU meant lower hosting costs for internal assistants and document processing.
Sources for this story
- Gemma 2 is now available to researchers and developers (opens in a new tab)Official
Google · blog.google
- gemma-2-27b model card (opens in a new tab)Official
Google on Hugging Face · huggingface.co
Last verified: . IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.
- #llm
- #gemma
- #open-weights
