For a deeper dive in the technical details Behind these capacities, as well as a complete overview of our approach to responsible development, refer to the Gemma 3 technical report.
Rigorous safety protocols to build Gemma 3 in a responsible manner
We believe that open models require a careful risk assessment, and our approach balances innovation with safety – adapt the intensity of tests to model capabilities. The development of Gemma 3 included in -depth governance of data, alignment with our security policies via fine and robust reference assessments. Although in -depth tests of more capable models often inform our less capable assessment, the improved STEM performance of Gemma 3 have caused specific evaluations focused on its abusive potential for the creation of harmful substances; Their results indicate a low risk level.
While the industry develops more powerful models, it will be essential to collectively develop approaches proportionate to security. We will continue to learn and refine our safety practices for models open over time.
Integrated security for image applications with Shieldgemma 2
Alongside Gemma 3, we also launch Shieldgemma 2A powerful 4B image safety verifier built on the Gemma 3. Shieldgemma 2 Foundation provides a ready -to -use solution for image safety, the release of safety labels in three security categories: dangerous content, sexually explicit and violence. Developers can personalize Shieldgemma more for their safety needs and users. Shieldgemma 2 is open and built to give flexibility and control, taking advantage of the performance and efficiency of the Gemma 3 architecture to promote responsible development of AI.
Ready to integrate the tools you already use
Gemma 3 and Shieldgemma 2 integrate perfectly with your existing workflows:
- Develop with your favorite tools: With support for the embraces facing Transformers, Ollama, JaxKeras, Pytorch, Google Ai Edge,, UnusualVLLM and Gemma.cppYou have the flexibility to choose the best tools for your project.
- Start experimenting in seconds: Get instant access to Gemma 3 and start building immediately. Explore its full potential in Google Ai Studioor download the models via Bother Or Face.
- Personalize Gemma 3 to your specific needs: Gemma 3 sent with a Reworked code base This includes recipes for effective adjustment and inference. Train and adapt the model using your favorite platform, as Google Colar,, Vertex ai Or even your game GPU.
- Deploy your way: Gemma 3 offers several deployment options, including Vetex AI, Cloud Run, the Google Genai API, IOCAUX environments and other platforms, offering you the flexibility to choose the best adjustment for your application and your infrastructure.
- Optimized performance experience on NVIDIA GPUs: NVIDIA has directly optimized the Gemma 3 models to make sure you get maximum performance on all -sized GPUs, from Jetson Nano to the last Blackwell chips. Gemma 3 is now presented on the API NVIDIA catalogallowing rapid prototyping with just an API call.
- Accelerate your AI development on many hardware platforms: Gemma 3 is also optimized for Google Cloud Tpus and integrates with AMD GPUs via the Open-Source Rocm ™ battery. For the execution of the processor, Gemma.cpp offers a direct solution.
A “gemverse” of models and tools
Gemverse is a vast ecosystem of Gemma models and tools created by the community, ready to feed and inspire your innovation. For example, Ai Singapore’s Sea-Lion V3 Decomposes linguistic barriers and promotes communication through Southeast Asia; BGGPT D’IN is a pioneering model of the great Bulgarian language which demonstrates the power of Gemma to support various languages; And Omniodio of Nexa Ai Presentation of the potential of the AI on disk, providing advanced audio treatment capacities to daily devices.
To promote more university research, we launch the Gemma 3 academic program. University researchers can request Google Cloud credits (worth $ 10,000 per allocation) to accelerate their research based on Gemma 3. The request form opens today and will remain open for four weeks. Apply Our website.
Start with Gemma 3
As part of our continuous commitment to democratize access to high quality AI, Gemma 3 represents the next step. Ready to explore Gemma 3? Here is where to start:
Instant exploration:
- Try Gemma 3 in full precision directly in your browser – no necessary configuration – With Google AI Studio.
- Get an API key directly from Google AI Studio and use Gemma 3 with the Google Genai SDK.
Personalize and build:
- Download Gemma 3 models from Face,, OlllaOr Bother.
- Easily refine and adapt the model to your unique requirements with the Hugging Face Transformers library or your favorite development environment.
Deployment and scale: