Understanding the language of single-cell biology
Based on Gemma and developed in collaboration with Yale University, Cell2Sentence-Scale 27B helped discover a potential new pathway for cancer therapy.
Welcome to the Gemmaverse
Join us in building AI that advances science and improves people’s lives.
Slide 1 of 3
Towards interspecies communication
Researchers at the Wild Dolphin Project and Georgia Tech treat dolphin vocalizations as a learnable language that can be tokenized, analyzed, and generated—just like human speech—with DolphinGemma.
Developing for Indic languages
Despite the incredible advances of AI, challenges remain in ensuring equitable representation and performance across diverse languages. Watch how developers at Telugu-LLM-Labs used Gemma to create Navarasa.
Trusted by organizations
Governments and public institutions enhance services and foster innovation by developing AI with Gemma.
Slide 1 of 3
The Ministry of Economy, Ecology and Agriculture of Ukraine digitizes licensing processes with Gemma
INSAIT created BgGPT, an open model for Bulgarian conversational and instruction-based tasks
GAIA is a collaboration between academia and industry to advance AI in Brazilian Portuguese
Loved by developers
Explore more open models in the Gemmaverse.
Slide 1 of 11
Lettria fine-tunes Gemma 3 for reliable knowledge graph creation
Swahili Gemma 1B from Crane AI Labs bridges the gap with powerful on-device performance
Adaptive ML trains Gemma 3 for exceptional multilingual results
Gemma 3 4B helps Levro build a model training platform with lower inference costs
SCBX Group improves the Thai language proficiency of its Typhoon model based on Gemma 3 12B
Nexa AI built its OmniAudio generative AI model for edge applications using Gemma
Sarvam AI built a translation model to handle all 22 officially recognized Indian languages
Quarks improves user experiences with Gemma 2 and Gemma 3
Roboflow improves computer vision with PaliGemma 2
AI Singapore makes AI more inclusive for Southeast Asia with Gemma 2
Institute of Science Tokyo creates powerful Japanese-focused LLM with Gemma 2
Built responsibly
Open models and tools are developed in accordance with AI Principles.
Slide 1 of 3
Safety
ShieldGemma 2 works as a defensive filter that sits between the user and the model to intercept and block toxic inputs or outputs.
Privacy
VaultGemma is trained with a differential privacy guarantee to ensure it learns new concepts without ever memorizing or leaking sensitive information.
Interpretability
Gemma Scope acts as a microscope for the model's inner workings using sparse autoencoders to visualize exactly how and why it makes specific decisions.