Browse Models
Gemma 3 is a family of open-weight multimodal large language models developed by Google DeepMind, featuring decoder-only transformer architectures with grouped-query attention. The family includes models ranging from 1B to 27B parameters, supporting text, vision, audio, and video modalities with context windows up to 128,000 tokens. Notable variants include the Gemma 3n E2B and E4B models that implement MatFormer architecture for efficient edge deployment and flexible parameter activation.