T5 Small

T5 Small

Turns one text into another with a small 60M-parameter T5 model that treats translation, summarization, question answering and classification as text-to-text jobs. It suits developers and students who want a small starting model.

Made by google-t5 on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: T5 Small is the 60 million parameter checkpoint of Google's Text-To-Text Transfer Transformer. Every task is cast as text in and text out, so one model, loss and set of settings serve many jobs.

The Hugging Face record tags it for translation and lists English, French, Romanian and German. Its builders also name summarization, question answering and classification as tasks the approach covers.

Who it suits: Developers and students who want a small, well-known language model to try or fine-tune suit it best. It runs from Python through transformers.

Industry | Who uses it | How they use it | Example
Software | Developers | Add small translation or text-to-text steps | Translate a short sentence between supported languages
Research | NLP researchers | Use it as a light baseline | Fine-tune it on a task of your own
Education | Students | Learn the text-to-text idea | Load it and try different inputs

Anyone who needs top quality or other languages can skip it, because this is the smallest checkpoint of the family.

Where to use it: You run it in Python on your own computer with the transformers library.

• Windows, macOS or Linux with Python installed
• Python scripts that use the transformers library
• Fine-tuning projects, since many people have published tuned versions

Use case | Who | What they get back
Translation | Developers | Text in another supported language
Summarization | Researchers | A shorter version of a text
Classification as text | Students | A label returned as a text string

How to set it up: You need Python and the transformers library. The card is short on setup, and the model files are small.

• Python on Windows, macOS or Linux
• The transformers library
• Disk space for about 242 MB of weights in F32

1. Install Python.
2. Install the transformers library with pip.
3. Download the model files with the hub tools.
4. Load the model by name in a Python script.

pip install -U huggingface_hub
hf download google-t5/t5-small

How to use it: The card text available here shows no run commands, and it points to its blog post and docs for details. You load the model like any transformers text-to-text model and give it a text string.

Example: You give the model a short English sentence to translate into German. It returns the German sentence as a text string.

For newcomers and experts: Think of the model as a text converter that takes one string and returns another. Your first small try is to download the files and translate one short sentence.

Under the hood, T5 turns each language task into text in and text out, with one model and one loss for all of them. It was pre-trained on the C4 corpus in a multi-task mixture of unsupervised and supervised tasks.

A task and its input as one text string
The T5 Small model reads it
It writes the answer as a text string

Good to know: The card marks its sections on risks, out-of-scope use and recommendations as more information needed.

• Size: 60.5M parameters, about 242 MB in F32, stored as Safetensors
• Languages: English, French, Romanian and German
• Training data: C4 and Wiki-DPR for the unsupervised part, plus supervised sets such as CoLA, SST-2 and MNLI
• Other versions: Hugging Face lists 21 quantized and 2,342 fine-tuned versions made by other people
• License: apache-2.0
• Cost to run: none named
• Data it reaches: the text you give it, and the model files downloaded from Hugging Face

On Hugging Face, 14 November 2026

• Task: translation
• Library: transformers
• Parameters: 60.5M
• Weights: about 242 MB in F32; about 30 MB at 4-bit (estimate)
• Formats: Safetensors
• Downloads in the last 30 days: 20,046,141
• Likes: 696
• License: apache-2.0
• Languages: en, fr, ro, de
• Quantized versions: 23
• Fine-tunes: 2,344
• Tags: transformers, pytorch, rust, onnx, safetensors, t5, text2text-generation, summarization, translation, multilingual

Links

• Model: https://huggingface.co/google-t5/t5-small
• Files: https://huggingface.co/google-t5/t5-small/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download google-t5/t5-small

Then follow the usage in its model card. Nothing in this kit runs by itself.