Google Gemma 4 Open Source: 31B Model Ranks #3, Apache 2.0, Full Agentic AI

# Google Gemma 4 Open Source: 31B Model Ranks #3, Apache 2.0, Full Agentic AI

> **Quick answer:** Google released Gemma 4 on April 2, 2026, under the Apache 2.0 license — the first time Google has offered a frontier-class open model with zero commercial restrictions. The 31B dense variant ranks #3 on the Arena AI text leaderboard. All four model sizes include native function-calling, structured JSON output, and multimodal processing, making them production-ready for agentic AI pipelines on your own hardware.

Google Gemma 4 open source model release under Apache 2.0 changes the calculus for every developer currently paying OpenAI or Anthropic API fees. Released April 2, 2026, Gemma 4 is not a research toy or a demo checkpoint — the 31B model scores 84.3% on GPQA Diamond and 80.0% on LiveCodeBench v6, benchmarks where models three to four times its size routinely underperform it.

## What Is Google Gemma 4 and What Did Google Actually Release?

Gemma 4 is Google DeepMind's fourth-generation family of open-weight language models. Unlike earlier Gemma releases — which carried usage restrictions that blocked commercial deployment at scale — Gemma 4 ships under the OSI-approved Apache 2.0 license. That single licensing change is arguably the biggest news in this release, and developer Sam Witteveen captured the significance plainly: "This is an actual real Apache 2 license, which means for the first time, you can take Google's best open model, modify it, fine tune it, deploy it commercially, do whatever you want with it."

Google released four model variants: