Launched in November 2021, Cedille’s mission is to make AI and Machine Learning for text generation simple and available to as many users as possible. With Cedille, entrepreneurs and large companies can leverage Machine Learning to write texts, summarize reports, classify contents or go next-level with SEO in French and German.

Supported by the Google TRC program, trained on TPUs and fully open-source, the language models were developed by ML engineers specialized in NLP at Coteries, a digital agency in Lausanne (CH). The multi-disciplinary team with strong ML competences released these models to provide better performance to professionals with single language use cases.

The French model of Cedille is 4 times larger than the current biggest existing French language model. With 6 billions parameters, Cedille can generate texts with unprecedented quality and can compete with multilingual models such as GPT-3.

The ML team at Coteries also strives to offer models that are free of potentially unsafe content and has taken special care to filter the data Cedille models are trained on. This process was made possible by a combination of large-scale data filtering using machine learning models to detect toxicity and unwanted content. On average, Cedille’s output is 14.7% safer compared to the best openly available system (GPT-fr). Complete benchmarking data and a research article with detailed performance data can be found on https://cedille.ai/. 

The team is currently working on new ways to make it even easier for customers to integrate Cedille into their products. Doing so will further lower the entry bar for anyone to use AI in their workflows and has the potential to unleash the power of natural language generation in a vast number of business use cases.

To date, more than 750’000 contents have been generated by 20’000 users. Try the models now on https://app.cedille.ai or integrate it in your processes with our API!