Google announces open Gemma 4 model with Apache 2.0 license

Google announced Gemma 4 as its latest open model; Gemma 4 is built from the same world-class research and technology as Gemini 3.

Available in four sizes: 31B Dense, 26B MoE, Effective 4B (E4B), and Effective 2B (E2B). The 2B and 4B models were developed with Pixel team, Qualcomm, and MediaTek.

Can run on phones, Raspberry Pi, and Jetson Nano with near-zero latency. Edge models have a 128K context window; larger offerings support up to 256K.

All models natively process video and images (multimodal). E2B and E4B provide native audio input for speech recognition and understanding. Trained on 140+ languages; includes advanced reasoning, agentic workflows, and code generation.

Released under the Apache 2.0 license (commercial/business-friendly). Weights are available on Hugging Face, Kaggle, and Ollama; access also via Google AI Studio and Google AI Edge Gallery.