Supertext

Supertext is a Swiss language AI provider offering secure and customisable AI translation in 28 languages.

Supertext is a Swiss language AI provider offering secure and customisable AI translation in 28 languages. All its systems are developed and hosted exclusively in Switzerland. Supertext is supporting the Swiss {ai} Weeks with free API access for all hackathons.

The Supertext API, launched in July 2025, allows users to automatically translate text and files in 28 languages - fast, reliable and secure. And it can do even more in its paid version. With a Supertext API subscription you can translate content from any third-party system automatically - and bring in experienced linguists to verify the AI output in minutes. Our goal is to create an all-in-one language workspace, where each intervention by professional linguists further improves and customises any user’s personal AI translation system.

Full documentation ...

Special offer for participants of the {ai} Hackathon:

👉 https://supertext.ubpages.com/api-swiss-ai-weeks/

Evaluating Supertext

The code below is from a 2025 study that compares DeepL and Supertext by assessing their performance on unsegmented texts. The authors include Alex Flückiger, Chantal Amrhein, Tim Graf, Frédéric Odermatt, Martin Pömsl, Philippe Schläpfer, Florian Schottmann, Samuel Läubli - open access

A/B Test Supertext vs DeepL

We release all evaluation data and scripts for further analysis and reproduction of the accompanying paper: A comparison of translation performance between DeepL and Supertext. The dataset is available on Hugging Face as well.

Installation

pip install poetry
poetry install

A/B Test Evaluation

To evaluate A/B results, call the script as follows:

poetry run python analysis/analyze.py -i data/ab_LANGPAIR.csv

You will find two TSVs with results in the results folder:

FILENAME_by_segment_winner: Aggregated results of segment wins by system
FILENAME_by_document_winner: Aggregated results of document wins by system

Citation

If you use our code or data, please cite our paper:

@misc{flueckiger-etal-2025-comparison,
    title={A comparison of translation performance between DeepL and Supertext}, 
    author={Alex Flückiger and Chantal Amrhein and Tim Graf and Frédéric Odermatt and Martin Pömsl and Philippe Schläpfer and Florian Schottmann and Samuel Läubli},
    year={2025},
    eprint={2502.02577},
    archivePrefix={arXiv},
    primaryClass={cs.CL},
    url={https://arxiv.org/abs/2502.02577}, 
}

👋 Contact ✨ Demo 💻 Source

Terms of Reuse