ai/gemma3n repository overview

⁠Gemma 3n

Gemma 3n is a compact, multimodal AI model from Google DeepMind, designed for efficiency on low-resource devices. It supports text, image, audio, and video input, with open weights and support for over 140 languages. With optimized parameter usage and strong safety features, it builds on the Gemma family to extend lightweight, high-performance foundation models.

⁠Intended uses

Gemma 3n models are designed for:

• Text generation: Compose poems, code, scripts, summaries, and more.
• Conversational AI: Power virtual agents and customer service assistants.
• Text summarization: Generate concise overviews of articles, papers, or transcripts.
• Language learning: Provide grammar suggestions and writing feedback.
• Educational research: Assist with data analysis, question answering, and exploration of multilingual resources.

⁠Characteristics

⁠Available model variants

¹: VRAM estimated based on model characteristics.

⁠Use this AI model with Docker Model Runner

To run the model:

docker model pull ai/gemma3n

Then launch it:

docker model run ai/gemma3n

More details in the Docker Model Runner documentation⁠.

⁠Benchmark performance

⁠Links

• Gemma 3n Model Page⁠
• Responsible AI Toolkit⁠