Today, we’re introducing TranslateGemma, a new collection of open translation models built on Gemma 3, available in 4B, 12B, and 27B parameter sizes. It represents a significant step forward in open translation, helping people communicate across 55 languages, no matter where they are or what device they own.
By distilling the knowledge of our most advanced large models into compact, high-performance open models, we have created a suite where efficiency doesn’t require a compromise on quality.
Outperforming models twice its size
The most remarkable finding in our technical evaluation is the efficiency of these models. Through our specialized training process, the 12B TranslateGemma model outperforms the Gemma 3 27B baseline as measured using MetricX on the WMT24++ benchmark.
For developers, this is a massive win. You can achieve high-fidelity translation quality using less than half the parameters of the baseline model. This efficiency breakthrough allows for higher throughput and lower latency without sacrificing accuracy. Similarly, the 4B model rivals the performance of the larger 12B baseline, making it a powerful model for mobile inference.
We tested TranslateGemma on the WMT24++ dataset, comprising 55 languages covering a wide variety of language families, including high-, mid- and low-resource languages. TranslateGemma considerably reduced the error rate compared to the baseline Gemma model in all languages, achieving improved quality with greater efficiency.
