Google Gemma 4

Open-weight multimodal AI models for developers and researchers

Visit Google Gemma 4 →

Gemma 4 is Google DeepMind's open-weight model family supporting text and multimodal inputs with advanced reasoning and agentic capabilities. It is designed to run efficiently across a range of hardware from mobile devices to GPUs, making it accessible for on-device and cloud deployments. Developers can use it to build AI applications with low compute overhead via Hugging Face, Ollama, or Google AI Studio.

At a glance

Company
Google DeepMind
Pricing
free
API available
Yes
Self-hostable
Yes
Launched
2026-04
Last verified
2026-05-11

Capabilities

multimodallong-contextfine-tuningon-device-inferenceagentic-workflowsopen-weights

Categories

Alternatives

For AI agents: machine-readable markdown version of this page at /tools/google-gemma-4.md, or send Accept: text/markdown.