A multilingual specialist built on Google's Gemma 27B foundation, this model focuses on translation tasks while also handling image inputs alongside text. It carries the heft of a 27-billion parameter architecture, which means it brings substantial language understanding to translation work, though that size also demands meaningful compute resources to run locally.