Google Gemma

Learn more in the blog blog.google/innovation-and…

Meet Gemma 4! Purpose-built for advanced reasoning and agentic workflows on the hardware you own, and released under an Apache 2.0 license. We listened to invaluable community feedback in developing these models. Here is what makes Gemma 4 our most capable open models yet: 👇

🌍 140+ Languages:Understanding of over 140 languages, Gemma 4 helps developers build inclusive, high-performance applications for a global audience.

📚 Longer context: Process large datasets seamlessly. Edge models feature a 128K context window, while larger models offer 256K.

✨Diverse and efficient architectures: Choose between Dense and Mixture-of-Experts (MoE) variants. The smaller models are specifically optimized for peak performance on mobile devices.

👁️👂 Vision + audio: Experience rich, multimodal interactions. All models handle images, with variable aspect ratio and resolution support, and video; smaller models feature native audio support for real-time edge processing.

💻Code Generation: Gemma 4 supports high-quality offline code, turning your workstation into a powerful, local-first AI code assistant.

🤖 Agentic Workflows: Native support for function-calling, structured JSON output, and native system instructions enables you to build agents that can interact with tools and APIs.

🧠 Advanced Reasoning: Gemma 4 demonstrates significant improvements in math and instruction-following tasks that require reasoning capabilities.

git commit -m "bump"

Two years ago, we released Gemma, Google DeepMind family of open models. Today, I'm thrilled to share a new milestone: Gemma 400M downloads and 100,000 variants! Thank you to every developer, partner, and contributor. We can't wait to see what you build next!👀

Introducing FunctionGemma Tuning Lab A deep dive guide + a no-code demo tool to fine-tune and export a model for your own, local-first tasks that require function calling. All powered by a 270m parameter model!

ollama run translategemma TranslateGemma is available on Ollama. Now you can use it in apps to translate between 55 languages. Note, it requires a specific prompting format 👇👇👇

TranslateGemma outperforms models twice its size, has been trained and evaluated on 55 language pairs to ensure high-quality performance, and uses a specialized two-stage fine-tuning process that distills the knowledge of Gemini models into an open architecture.

More Gemma! Meet TranslateGemma, a new collection of open translation models built on Gemma 3 designed for high-performance communication. - Available in 4B, 12B, and 27B parameter sizes. - Evaluated on 55 languages using the WMT24++ dataset. - 12B model outperforms the Gemma 3 27B baseline. - 4B model optimized specifically for mobile and edge deployment. huggingface.co/collections/go…

We’re releasing TranslateGemma, a new family of open translation models with support for 55 languages. 🌐 Available in 4B, 12B, and 27B parameter sizes – they’re designed for efficiency without sacrificing quality.

TranslateGemma Technical Report huggingface.co/papers/2601.09…

I'm excited to introduce the new addition to the Gemma family: TranslateGemma 🎉 - Trained on translation tasks across 55 language pairs - 4B, 12B, and 27B parameters - With multimodal input We're also publishing a technical report with an overview of training and evaluation