Models

← All models

T5 3B

Google DeepMind · 2019-10-23 · 2.9B parameters

T5-3B is the approximately 2.8-billion-parameter variant of Google's T5 (Text-to-Text Transfer Transformer) family, introduced in October 2019. T5 reformulates all NLP tasks as a unified text-to-text problem — feeding both inputs and target outputs as strings to an encoder-decoder Transformer — and was pre-trained at scale on a cleaned subset of Common Crawl (C4) to enable transfer learning across translation, summarization, classification, and question answering.

Benchmark scores