Gemma 4
Definition
Gemma 4 is a family of openly available Google models for generative and multimodal tasks.
Gemma 4 includes models with open weights from Google DeepMind. The model card mentions dense architectures and mixture-of-experts variants, where selected model parts contribute to processing. Supported input types and hardware requirements differ between variants.
Simply explained
Like blueprints for several machines: You can operate a suitable version yourself. An accessible blueprint does not mean every machine runs on every computer.
Why it matters in practice
For self-hosted model deployment, specific model size and required modalities are decisive. The Gemma 4 model card distinguishes variants for different device classes.
Example from business
Possible example: A development team tests a Gemma 4 variant on its own hardware to summarize released internal texts.
Limits and pitfalls
- Audio processing is not indicated for all variants. The model card also mentions possible biases and limits of generative outputs; a locally run variant also requires its own tests.
Origin & history
Google's release directory lists March 31, 2026, for the first listed Gemma 4 variants and June 3, 2026, for Gemma 4 12B Unified.
Use in mid-sized companies
- Possible applications are local text prototypes, image evaluation, and testing self-operated assistant systems.
Frequently asked questions
Are open weights the same as a hosted API service?
No; the model card describes deployed models for different operating environments.
Can all variants process audio?
The cited model card assigns this capability only to certain variants.
Sources
- Gemma 4 model card · Google DeepMind (2026): Gemma 4 model card
- Gemma releases · Google (2026): Gemma releases
Editorially reviewed on: