Google's Gemma open models pass 1 billion downloads

Google DeepMind's Gemma family of open models has surpassed 1 billion downloads, with developers publishing more than 100,000 variants over the past two years, the company said in a blog post covered by The Next Web.

"What matters far more than the download count is what the community is building with them," wrote Clement Farabet, Vice President at Google DeepMind, and Olivier Lacombe, a Product Director on the team. Google calls the ecosystem around those variants the Gemmaverse.

The headline number understates reach. Gus Martins, a Google developer relations engineer, flagged that the total "doesn't count Android and Chrome integrations." Omar Sanseviero from the Gemma team summed up the spread as "From running underwater to running at space." The showcase list runs from NASA, Satlyt and Starcloud putting Gemma in orbit for onboard image analysis, to India's Aarogya Setu 2.0 health app using it for medical report digitization, to a Yale and Google model called C2S-Scale probing novel cancer therapy pathways in living cells. The All India Institute of Medical Sciences is using MedGemma for outpatient triage; the same model is supporting health workers in rural Uganda.

The comparison Google did not lead with: Alibaba claimed 3 billion downloads for Qwen five days earlier, with more than 300,000 derivatives. Hugging Face's own report counted 2,045 million Qwen downloads for the year and 151,448 Qwen-based derivatives, about 2.6 times Meta's entire footprint. It slots into a busy run of open-model releases we've been tracking this quarter, with 266 Gemma-adjacent stories in our feed over the last 90 days.