For a deeper dive in the technical details behind these capabilities as well as a comprehensive overview of our approach to responsible development, see Gemma 3 technical report.
Strict Security Protocols To Build Gemma 3 Responsible
We believe that open models require careful risk assessment and our approach balances innovation with certainty – tailoring test intensity for model functions. Gemma 3’s development included extensive data management, adaptation to our security policies via fine -tuning and robust benchmark evaluations. While thorough testing of more skilled models often informs our assessment of less talented mother, Gemma 3’s improved voice evil gained that got specific evaluations that focused on its potential for abuse in creating harmful substances; Their results indicate a low risk level.
As the industry develops more powerful models, it will be critical to collectively develop risk -breaking approaches to safety. We will continue to learn and refine our security practice for open models over time.
Built -in security for image applications with Shieldgemma 2
Alongside Gemma 3 we also launch Shieldgemma 2, a powerful 4B image security checker built on the Gemma 3 Foundation. Shieldgemma 2 provides a ready -made image security solution that emits safety marks across three security categories: dangerous content, sexually explicit and violence. Developers can further customize shieldgemma to their security needs and users. Shieldgemma 2 is open and built to provide flexibility and control, utilizing the performance and efficiency of the Gemma 3 architecture to promote responsible AI development.
Ready to integrate with the tools you are already using
Gemma 3 and Shieldgemma 2 integrate seamlessly into your existing workflows:
- Develop with your favorite tools: With support to embrace face transformers, Ollama, Jax, Keras, Pytorch, Google Ai Kant, Unloth, Vllm and Gemma.cpp, you have the flexibility to choose the best tools for your project.
- Start experimenting in a few seconds: Get instant access to Gemma 3 and start building right away. Explore its full potential in Google AI Studio, or download the models through Kaggle or embraced face.
- Customize Gemma 3 for your specific needs: Gemma 3 vessels with a renewed code base that include recipes for effective fine tuning and inference. Train and customize the model using your favorite platform, such as Google Colab, Vertex AI or even your gaming GPU.
- Distribute your way: Gemma 3 offers several implementation options including Vertex AI, Cloud Run, Google Genai API, IOCAL environments and other platforms, giving you the flexibility to choose the best fit for your application and infrastructure.
- Experience Optimized Performance on NVIDIA GPUs: Nvidia has directly optimized Gemma 3 models to make sure you get maximum performance on GPUs of any size, from Jetson Nano to the latest Blackwell chips. Gemma 3 is now shown on the NVIDIA API catalog, enabling fast prototype with only an API call.
- Accelerates your AI development across many hardware platforms: Gemma 3 is also optimized for Google Cloud TPUs and integrated with AMD GPUs via Open Source ROCM ™ stack. For CPU performance, Gemma.cpp offers a direct solution.
A “gem maver” of models and tools
Gemmavers is a huge ecosystem of socially created Gemma models and tools, ready to power and inspire your innovation. For example, AI Singapore’s Sea-Lion V3 breaks down language barriers and promotes communication over Southeast Asia; Insaves BGGPT is a groundbreaking Bulgarian-first great language model that demonstrates Gemma’s power to support different languages; And Nexa AIS Omniaudio shows the potential for AI on devices, bringing advanced audio treatment features to everyday devices.
To further promote breakthroughs in academic research, we launch the Gemma 3 Academic Program. Academic researchers can apply for Google Cloud Credits (worth $ 10,000 per Award) to speed up their Gemma 3-based research. The application form opens today and remains open for four weeks. Apply to our website.
Get started with Gemma 3
As part of our ongoing commitment to democratizing access to AI in high quality, Gemma 3 represents the next step. Ready to explore Gemma 3? Here you have to start:
Instant Investigation:
- Try Gemma 3 at full precision directly in your browser – no setup needed – with Google AI Studio.
- Get an API key directly from Google AI Studio and use Gemma 3 with Google Genai SDK.
Customize and build up:
- Download Gemma 3 models from Hugging Face, Ollama or Kaggle.
- Easy to fine-tune and customize the model to your unique requirements with Hugging Face’s Transformer Library or your preferred development environment.
Deploy and scale: