Gemma 3:4b

Gemma 3 4B is Google DeepMind's 4-billion-parameter, multimodal open model (text and image) for efficient on-device and cloud deployment.

This is the 4B-parameter variant of the Gemma 3 family: a state-of-the-art, lightweight, open model from Google DeepMind, built on the same core technology as Gemini. It delivers multimodal capabilities (text and image) and supports an extensive 128K token context window. Optimized for efficiency, the 4B size is ideal for deployment on consumer hardware (laptops, mobile devices) and cloud services, offering robust performance for tasks like complex reasoning, summarization, and question answering across over 140 languages. The open weights promote extensive customization and responsible AI development.

https://ai.google.dev/gemma