Google released Gemma 4 on April 2, 2026, its most capable open-weight family to date under an Apache 2.0 license. The lineup spans four sizes — E2B and E4B for on-device deployment and a 26B MoE and 31B Dense for cloud workloads — all natively multimodal with 256K-token context, function calling, and 140+ language support. The E-series models run fully offline on phones, Raspberry Pi, and edge hardware in collaboration with Qualcomm and MediaTek; the larger variants are accessible via Google AI Studio and Vertex AI. Gemma models have been downloaded over 400 million times since the first generation.