By July 2026, Google DeepMind's Gemma family of open AI models had surpassed 900 million total downloads. The latest version, Gemma 4, alone accounted for more than 300 million of those since its launch in April 2026.
900 Million Downloads and Counting
The milestone marks a rapid climb for the Gemma family, which first debuted in early 2024. The models are designed to be lightweight and accessible, targeting developers and researchers who want to build AI applications without relying on massive proprietary systems. The 900 million figure includes downloads across all versions, from the original Gemma to the current Gemma 4.
Google DeepMind has not disclosed a breakdown by region or use case, but the sheer volume suggests broad adoption in both academic and commercial settings. Open-weight models like Gemma have become a popular alternative to closed-source offerings, allowing users to fine-tune and deploy them on their own infrastructure.
Gemma 4's Contribution
Gemma 4, released in April 2026, has been the fastest-growing version in the family's history. Its 300 million downloads in roughly three months indicate strong demand for the latest capabilities. The model builds on previous iterations with improved performance on reasoning and coding tasks, though Google DeepMind has not released detailed benchmarks for this version.
The company has positioned Gemma 4 as a tool for both experimentation and production use. Its open license allows for commercial applications, which likely contributed to the rapid uptake. Developers can access the model through Google's AI platform and other distribution channels.
The 900 million download figure was reported as of July 2026. Google DeepMind has not announced a timeline for the next Gemma release, leaving the community to watch for updates.
