<ul data-eligibleForWebStory="false"><li>Google has launched T5Gemma, a new collection of encoder-decoder large language models (LLMs) promising improved quality and efficiency compared to decoder-only models.</li><li>T5Gemma is based on the Gemma 2 framework and uses an adaptation technique to convert pretrained decoder-only models into encoder-decoder models.</li><li>Benchmark results show T5Gemma models outperform decoder-only models in accuracy and latency, dominating the quality-inference efficiency frontier.</li><li>Google has released a range of T5Gemma checkpoints for experimentation and deployment on platforms like Hugging Face, Kaggle, and Vertex AI.</li></ul>

Google Launches T5Gemma to Reclaim Encoder-Decoder Architecture Benefits

Discover more