Gemma 4

by Google · Website

Google's Gemma 4 is the most capable open model family from Google DeepMind, released April 2026 under Apache 2.0. It spans four sizes from edge (E2B, E4B) to frontier (26B MoE, 31B Dense), all supporting vision, 140+ languages, and up to 256K context. The 31B Dense model ranks #3 on Arena AI among open models, outcompeting models with 20x more parameters.