Skip to content

# Gemma 4: A Byte-Level Look at the Strongest Open-Source Model

· DeepMind Translated
DeepMind

Gemma 4: The strongest open model, down to every byte

Gemma 4 delivers our most powerful open model yet, with improved advanced reasoning, better handling of complex instructions, and support for multi-step workflows.

Built with native multimodal capabilities, Gemma 4 can understand both text and images and demonstrates strong performance across a range of benchmarks.

Key Features

  • Stronger reasoning capabilities
  • More reliable instruction following
  • Supports multimodal input
  • Optimized for agentic workflows

For Developers

Gemma 4 is a good fit for developers who want to build high-quality applications on open models. It is especially well suited for the following use cases:

  • Complex question answering
  • Content generation
  • Code assistance
  • Multi-step task execution
  • Applications that require image understanding

Safety and Responsible AI

Gemma 4 continues our commitment to safety and responsible AI, and aims to help developers build applications in a more controllable way.

Availability

Gemma 4 is now available to developers.