Google Releases Gemma 4 Under Apache 2.0 License

Google DeepMind released Gemma 4 on April 2, 2026, marking the largest open model release in the Gemma series since it launched in 2024. The family spans four sizes: E2B and E4B edge models optimized for phones, Raspberry Pi, and offline use, a 26B mixture-of-experts model that activates only 4B parameters during inference, and a 31B dense flagship that runs on a single 80GB GPU. All models are released under the Apache 2.0 license, a first for Gemma and a meaningful departure from the custom licensing that restricted earlier versions. The technical numbers are strong. Gemma 4 supports context windows up to 256,000 tokens, processes text, images, video, and audio natively, works across more than 140 languages, and is built for agentic workflows with native function calling and structured output. On competitive coding benchmarks, the 31B model scored an ELO of 2150 on Codeforces compared to 110 for Gemma 3, nearly a 20x improvement in one generation. The community has downloaded Gemma models more than 400 million times and built over 100,000 variants. Day-one support is available across Hugging Face, Ollama, vLLM, llama.cpp, MLX, LM Studio, and NVIDIA NIM. The Apache 2.0 license is the most important part of this release for developers and businesses. Previous Gemma versions carried custom terms that created friction for enterprise adoption and commercial products. Apache 2.0 removes all of that. Developers building SaaS products, enterprise tools, or local AI applications no longer face licensing risk when building on Gemma 4. For developers in Nigeria and across Africa, the edge models are directly relevant. Gemma 4 E2B and E4B run fully offline on Android devices, including billions of mid-range smartphones already in the market. Building AI features for users with inconsistent connectivity becomes much more practical when the model runs on-device without a cloud call. If you are building AI products and have not evaluated Gemma 4 yet, this release deserves a serious look.