Gemma 3 1B

A high-performance 1-billion parameter multimodal model designed for efficient local execution and cross-modal reasoning.

Gemma 3 1B delivers native multimodal capabilities to edge devices using the same architectural foundations as Google's Gemini models. This compact 1-billion parameter model processes text and image inputs simultaneously, enabling complex reasoning on consumer-grade hardware with minimal latency. It operates efficiently within tight memory constraints (typically under 2GB when quantized to 4-bit) and provides a versatile solution for real-time visual analysis or private local chat. Developers leverage its small footprint to deploy sophisticated AI features directly into mobile apps and browser-based environments.

https://blog.google/technology/google-deepmind/google-gemma-3/