T5Gemma 2

It helps you a family of open-source encoder-decoder models with 2 to 9 billion parameters, based on Gemma 2, with higher inference efficiency than decoder-only models.

Best for: developers and researchers building chatbots or language applications