Gemma 4 Multimodal Family Released with Tool Calling Capabilities

Google DeepMind released the Gemma 4 family of multimodal models on Hugging Face, licensed under Apache 2.0. These models support image, text, and audio inputs to generate text responses, offering frontier-level performance for reasoning, agentic workflows, and coding. Designed for various environments including on-device deployment, they feature a 256K-token context window, support over 140 languages, and include enhanced tool calling capabilities.