T5Gemma 2: The next generation of encoder-decoder models

Google released T5Gemma 2, an upgraded encoder-decoder model family that adds multimodal vision capabilities and 128K token context windows. The models incorporate tied embeddings and merged attention mechanisms for efficiency, while supporting over 140 languages. Available in compact sizes from 370M to 7B parameters, they outperform predecessor models on coding, reasoning, and long-context tasks.