T5

T5 is a 2019 Google model that treats every language task as turning one piece of text into another, from translation to summarising.

In plain words

T5 stands for Text-to-Text Transfer Transformer. Instead of building a different output for each task, it always takes text in and gives text out, with a short prefix saying what to do. That let Google's researchers compare many ways of pre-training and fine-tuning in one framework. They pre-trained T5 on C4, a large cleaned-up dataset built from Common Crawl web pages, and released the data, models and code.

A quick example

To translate, you give T5 a German sentence that starts with a short instruction asking for English. It replies with the English sentence as plain text. For other jobs, only the instruction at the front changes.

Where you'll meet it

• Translation, summarisation and question answering research.
• A text encoder inside other models, such as the Imagen image model.

Related

Sources

1. PaperExploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer, Raffel et al. (arXiv) · read 29 Sept 2026
2. Docstext-to-text-transfer-transformer, Google Research (GitHub) · read 12 Nov 2026
3. PaperPhotorealistic Text-to-Image Diffusion Models with Deep Language Understanding, Saharia et al. (arXiv) · read 29 Sept 2026
4. DocsExploring Transfer Learning with T5: the Text-To-Text Transfer Transformer, Google Research · read 29 Sept 2026

Checked by WhatIsAI against the sources above · 12 Nov 2026. Spotted an error? Report it.