Google’s Gemma 4 AI models get 3x speed boost by predicting future tokens

Up to 3x the speed with no loss of quality—is it too good to be true?