Google says its Gemma family of open models has passed a billion downloads.<\/p>\n
Developers have published more than 100,000 Gemma variants over the past two years, the company said in a blog post<\/a> on Thursday. Clement Farabet, a vice president at Google DeepMind, and Olivier Lacombe, a product director there, wrote it.<\/p>\n Google calls the ecosystem around those variants the Gemmaverse.<\/p>\n \u201cWhat matters far more than the download count is what the community is building with them,\u201d the two wrote.<\/p>\n What the number counts<\/p>\n Gus Martins, a Google developer relations engineer, added a caveat on Bluesky.<\/p>\n \u201c1B Gemma downloads,\u201d he wrote. \u201cFun fact: this doesn\u2019t count Android and Chrome integrations.\u201d<\/p>\n Google did not break the total down by model, by year, or by platform.<\/p>\n Omar Sanseviero, who works on Gemma at Google DeepMind, described the spread on X. \u201cFrom running underwater to running at space,\u201d he wrote.<\/p>\n How that compares to Qwen<\/p>\n Alibaba made a larger claim five days earlier.<\/p>\n TNW reported on 16 August that Alibaba said Qwen had passed 3 billion downloads<\/a>. Hugging Face\u2019s State of Open Models report counted 2,045 million Qwen downloads this year.<\/p>\n On derivatives, Alibaba claimed more than 300,000. Hugging Face counted 151,448 Qwen-based derivatives on its hub, which the report said was 2.6 times Meta\u2019s entire footprint there.<\/p>\n Google\u2019s figure of 100,000 Gemma variants covers two years and is the company\u2019s own count.<\/p>\n Hugging Face says its download figures measure activity inside its own hub. They do not capture API usage, private deployments or models distributed elsewhere.<\/p>\n Gemma in orbit<\/p>\n Google says NASA, Satlyt and Starcloud are running Gemma directly in orbit.<\/p>\n The uses it lists are onboard image analysis, optimising scarce downlink bandwidth and routing communications between satellites. Satlyt and Starcloud are both startups; NASA is the agency.<\/p>\n Google says those deployments show Gemma can handle complex reasoning in constrained environments.<\/p>\n It does not say which Gemma models are running on which spacecraft, or since when.<\/p>\n Gemma in India\u2019s health app<\/p>\n India\u2019s National Health Authority has integrated Gemma 4 and Google\u2019s open-source Medical Data Toolkit into Aarogya Setu 2.0.<\/p>\n Google says that app has over 100 million downloads on Android. Gemma 4 turns medical reports into standardised digital formats, which Google says lets citizens manage and share health records across providers.<\/p>\n The cancer pathway claim<\/p>\n Researchers from Yale and Google built a model called C2S-Scale on Gemma, designed to interpret the language of single cells.<\/p>\n Google says C2S-Scale \u201csuccessfully discovered a novel cancer therapy pathway that was verified in living cells\u201d. It calls that the first time an AI system produced novel mechanistic therapeutic pathways verified in living cells.<\/p>\n The claim appears in Google\u2019s own announcement. In this post the company links to its earlier write-up of the work rather than to a paper.<\/p>\n MedGemma and DolphinGemma<\/p>\n MedGemma is Google\u2019s domain-specific medical model.<\/p>\n Google says organisations are using it in clinical applications. The examples it gives are outpatient triage at the All India Institute of Medical Sciences, and systems supporting frontline health workers in rural Uganda.<\/p>\n DolphinGemma came out of a collaboration with Georgia Tech and the Wild Dolphin Project. Google says the model processes dolphin vocalisations to predict sound sequences. It describes the work as ongoing and says researchers are still iterating.<\/p>\n The Awesome Gemma repository<\/p>\n Google has launched a GitHub repository called Awesome Gemma as the official directory for the Gemmaverse. It says the repository will feature community projects, fine-tunes, tutorials and developer tools.<\/p>\n Philipp Schmid, who works on Google\u2019s AI developer experience, listed the contents on X. Model cards and collections for 16 Gemma variants, setup guides for Ollama, vLLM and LiteRT, and fine-tuning recipes for Unsloth, Tunix and MLX.<\/p>\n TNW reported in July that Ollama raised $65mn<\/a> as its open-model runner reached nearly 9 million developers.<\/p>\n Google also says its Gemma Challenge on Kaggle drew more than 1,600 project submissions, with winners to be announced.<\/p>\n Ollama posted its own note. \u201cGemma is one of the most popular open models,\u201d the company wrote. \u201cIt\u2019s been an amazing journey being a close partner.\u201d<\/p>\n Where Gemma runs<\/p>\n Google says Gemma was built to deploy across environments, from local devices to edge infrastructure.<\/p>\n TNW reported this month that Google built an offline translator<\/a> running on a Raspberry Pi using Gemma.<\/p>\n Lacombe described the original goal on X. \u201cBuild open models that are lightweight, highly capable, and flexible enough to run anywhere,\u201d he wrote.<\/p>\n In last year\u2019s equivalent post, Google highlighted developers building AI assistants for visually impaired people and offline educational hubs for regions without connectivity.<\/p>\n The open-model field around it<\/p>\n Other labs have been shipping open models through the same period.<\/p>\n Nvidia is working towards a trillion-parameter open model<\/a>, TNW reported this month. Nvidia has said it is chasing China on open weights.<\/p>\n