Google has officially unveiled Gemma 4, a cutting-edge family of four open-weight models engineered to span a wide range of computational environments, from compact smartphones to powerful workstations. These models draw directly from the same foundational research that powered the acclaimed Gemini 3. The new suite includes a highly efficient 2B edge model, specifically designed to run on low-power devices such as a Raspberry Pi, demonstrating its versatility. At the higher end, it features a robust 31B dense model, which has already secured the third position on the respected Arena AI open-model leaderboard for its performance. A significant development with this release is the adoption of the Apache 2.0 license, representing a notable and more open approach compared to previous Gemma iterations.
