Google DeepMind says its Gemma family of open models has surpassed 1B downloads and developers have published 100K+ Gemma model variants over the past two years (Google)

Google's DeepMind has revealed that its Gemma family of open models has achieved a significant milestone, surpassing one billion downloads. Since its introduction, this versatile AI model has been deployed across various environments, from local devices to edge infrastructure and even space. The Gemma family boasts over one-hundred thousand model variants built by developers, forming a thriving ecosystem known as the Gemmaverse.

Notably, the Gemma models have been instrumental in powering onboard image analysis for NASA, optimizing downlink bandwidth for Satlyt and Starcloud, and streamlining complex medical reports for India's National Health Authority. Moreover, researchers from Yale University and Google have developed C2S-Scale, an AI model based on Gemma that has discovered a new cancer therapy pathway verified in living cells.

The Gemmaverse demonstrates its versatility by aiding in the development of healthcare applications, such as improving workflow management and patient communication, and supporting frontline health workers in rural Uganda. The platform has also extended under the sea, with the creation of DolphinGemma, an AI model that processes dolphin vocalizations to predict sound sequences.

The community's impact is further highlighted by the Gemma Challenge on Kaggle, which has seen over 1600 projects submitted to address real-world challenges. To further facilitate collaboration and knowledge sharing, Google has launched the Awesome Gemma repository on GitHub, serving as an official directory of the Gemmaverse, featuring the best community projects, fine-tunes, tutorials, and developer tools.