Gemma-fied! Putting the power of our most capable family of open models into the hands of every Indian developer
Our Gemma family of open models demonstrates our commitment to making useful AI technology accessible to all, leveraging the same technology that powers our Gemini models. The Gemma-powered innovations Indian developers have been building – leveraging its multimodal, long context and multilingual capabilities – embody this commitment. Our new Gemma 3 models are the best multilingual open models on the planet, supporting over 140 languages, including six Indian languages.
We're pushing that boundary even further with specialized Gemma variants like ShieldGemma for safer LLM deployments, and MedGemma, our most capable open multimodal model for building health applications for which we released new versions just two weeks ago.
And the spirit of open innovation even extends to groundbreaking research, with scientists using DolphinGemma, the world’s first large language model for dolphins – imagine, one day, having a conversation with a dolphin!
Inclusivity is also about bringing powerful AI directly to users. At I/O, we introduced Gemma 3n, our latest fast and efficient open multimodal model engineered to deliver highly capable, real-time AI directly on-device. It is capable of running on phones, laptops, and tablets with as little as 2 gigabytes of RAM.
Gemma 3n is designed to enable developers to create multimodal AI experiences that are fast, private, and available even offline, with audio capabilities including high-quality automatic speech recognition and translation. And with a low footprint, solutions built on Gemma 3n can run smoothly on phones across the spectrum of capacity and affordability, bringing AI-powered innovations closer to every Indian.
In the last few weeks, we have begun rolling out the availability of Gemma 3n to Indian developers and the response has been amazing.
We’re also thrilled to share today that 3 startups selected by the India AI Mission - Sarvam, Soket AI and Gnani - are building the next generation of Make in India AI models on Gemma. We are also closely collaborating with Sarvam, which has already built a long-form text translation model using Gemma.
We are also closely partnering with BharatGen at IIT Bombay to build indigenous Indic language Automatic Speech Recognition (ASR) and Text-to-Speech (TTS) models that will aim to make Gemma more helpful in the Indic context, especially for historically underrepresented languages and governance applications.
By democratizing access to these powerful, localized AI models and tools, we're aiming to not just support developers, but empower every Indian to benefit wholly from the AI revolution, with solutions that truly reflect the nation's remarkable diversity.
