Alphabet’s Gemma passes one billion downloads as Google spotlights edge and science use cases

Google is marking a major milestone for its open AI model strategy, saying the Gemma family of models has surpassed one billion downloads. In a post published Tuesday, the company framed the number as a community announcement while emphasizing what developers have built using the models, including an expanding ecosystem of model variants and applications aimed at real-world problems.

When Google introduced Gemma, it said its goal was to help developers build responsible AI applications across a range of environments, from local devices to edge infrastructure and, in one example, space. Google’s latest update credits that flexibility for adoption by teams building in constrained settings where compute and bandwidth are limited.

Google said that over the past two years developers have published more than 100,000 Gemma model variants. The company described the resulting body of work as the “Gemmaverse,” a term it uses for the community ecosystem forming around Gemma and related tools.

One of the featured case studies involves space-based computing. Google said teams at NASA, Satlyt, and Starcloud are running Gemma directly in orbit for tasks including onboard image analysis, optimizing scarce downlink bandwidth, and routing intersatellite communications. Google said the point of these projects is that Gemma can be used for complex reasoning in extreme, resource-constrained environments.

Back on Earth, Google highlighted a health-data deployment in India. The company said India’s National Health Authority integrated Gemma 4 and Google’s open-source Medical Data Toolkit into Aarogya Setu 2.0, an Android app with more than 100 million downloads. Google said the app uses Gemma 4 to process complex medical reports into standardized digital formats intended to help citizens manage and securely share health data across providers.

In research, Google described an effort led by researchers from Yale and Google to build C2S-Scale, an AI model designed to interpret the “language” of single cells. Google said C2S-Scale is built on Gemma and that it identified a novel cancer therapy pathway, which the company says was verified in living cells. Google also characterized this as the first time an AI system produced novel mechanistic therapeutic pathways that were verified in living cells, a claim it attributes to the described research.

Google also cited domain-specific work it says is designed to speed healthcare application development. The company referenced MedGemma, saying it helps developers accelerate building healthcare tools, including workflow management, patient communication, and decision support for diagnostics and treatment. Google said MedGemma is being used by organizations globally, including for outpatient triage support at AIIMS (All India Institute of Medical Sciences) and for systems supporting frontline health workers in rural Uganda, without offering additional performance details.

Beyond terrestrial and space deployments, Google pointed to a marine-focused project. In collaboration with Georgia Tech and the Wild Dolphin Project, Google said researchers developed DolphinGemma, a model intended to process dolphin vocalizations to predict sound sequences. Google said researchers are still iterating on the challenge and suggested that more work is coming from underwater applications.

Looking beyond individual examples, Google tied its one-billion-downloads milestone to community participation. It said last year it highlighted developers building a range of tools, from AI assistants for people with visual impairments to offline educational hubs for disconnected regions. This momentum, the company said, continued through a Gemma Challenge on Kaggle, where the community submitted more than 1,600 projects aimed at solving real-world problems, with winners to be announced soon.

To give developers a way to share and discover Gemma-related work, Google said it is launching the Awesome Gemma repository on GitHub. The company described it as a curated directory intended to serve as an official hub for the Gemmaverse, featuring community projects, fine-tunes, tutorials, and developer tools. Google said the repository is meant to help people find projects and start collaborating, positioning it as a “launchpad” for builders at different experience levels.