Gemma 4, Google's latest generation of open multimodal models, is now available on Docker Hub as OCI artifacts. The lineup includes four model variants ranging from small efficient models (E2B at 5.1B params, E4B at 8B params) to a Mixture-of-Experts model (26B A4B) and a flagship dense model (31B) with up to 512K context windows. All models support text, vision, and in some cases audio inputs. Docker packages them like containers, enabling a simple `docker model pull gemma4` command with no custom toolchains. Integration with Docker Model Runner for Docker Desktop is coming soon, allowing users to run and manage models directly from their local environment.

3m read timeFrom docker.com
Post cover image
Table of contents
What Docker Brings to Gemma 4What’s New in Gemma 4?Technical SpecificationsBuild the Future of AI with Docker Hub

Sort: