Google DeepMind has launched Gemma 4, its most capable open-weight AI model family to date, engineered to run complex reasoning tasks on low-power devices like smartphones and workstations.
Built on the same foundation as Gemini 3, Gemma 4 delivers unprecedented intelligence per parameter. The 31B Dense variant ranks third among open models on the Arena AI Text leaderboard.
The lineup includes four variants: Effective 2B, Effective 4B, a 26B Mixture of Experts model, and a 31B Dense model. The smaller E2B and E4B models support native audio input, enabling real-time speech recognition directly on device.
All models feature a 128K to 256K context window, allowing users to process entire codebases or massive document sets in a single prompt. Native function calling and structured JSON output enable autonomous AI agents to interact with external tools without custom coding.
Each model is released under the permissive Apache 2.0 license, removing commercial barriers for enterprise developers.
Analyst Holger Mueller of Constellation Research noted, "Google is building its lead in AI not just through Gemini, but by dominating the local AI ecosystem with Gemma 4. These models enable functional, vertical applications across device form factors."
With GPU-compatible sizes and edge-first design, Gemma 4 positions Google at the forefront of decentralized, sovereign AI deployment.