Google has unveiled TranslateGemma, a new family of open-source language models specifically designed for high-quality translation. Built upon the Gemma 2 architecture, these compact yet powerful models (2B and 7B parameters) support over 100 languages and set new benchmarks for open-source machine translation, making state-of-the-art translation technology more accessible to developers and researchers worldwide.

Google has officially released TranslateGemma, a new suite of open-source machine translation models built upon the Gemma 3 architecture. Available in 4B, 12B, and 27B parameter sizes, these models deliver state-of-the-art translation quality for 55 language pairs while achieving a remarkable efficiency breakthrough. The 12B variant notably outperforms the baseline Gemma 3 27B model, offering high-fidelity translation with less than half the parameters