AI Model

Gemma 2

Released on June 27, 2024

Changes from previous version

New architecture generation with 9B and 27B models; 2B followed in July. Large gains on MMLU and coding versus Gemma 1 at similar serving cost.

Release Summary

Gemma 2 in 9B and 27B sizes with grouped-query attention and stronger efficiency at inference than Gemma 1.

Timeline

July 31, 2024

Gemma 2 2B released

Google adds a 2B Gemma 2 checkpoint for edge and mobile deployments.

June 27, 2024

Gemma 2 (9B, 27B) released

Google launches Gemma 2 flagship open weights in 9B and 27B sizes.