Google's Gemma 4 is now available with Apache 2.0 licensing for the first time

Google released Gemma 4, its most capable open model family yet, published under a full Apache 2.0 license for the first time.

Gemma 4 is based on the same technology as Google's proprietary Gemini 3.

It brings significant improvements in multi-step reasoning and math, and natively supports function calling, structured JSON output, and system instructions.

The family includes four sizes—E2B (effective 2B), E4B (effective 4B), 26B Mixture-of-Experts, and 31B Dense—targeting devices from smartphones and Raspberry Pi to workstations (including Jetson Orin Nano).

The 31B Dense uses unquantized bfloat16 weights that fit on a single 80 GB NVIDIA H100, while quantized versions should run on consumer graphics cards.