Gemma 3 4B

Google's compact open-weight model with 4 billion parameters from the Gemma 3 family. Supports multimodal inputs including text and images with a 128K context window. Balances efficiency and capability for vision and language tasks.