DeepLThe company recently announced the launch of its next-generation language model, claiming that the model has surpassed Google Translate and GPT-4 in translation accuracy. The new generation of language models uses more than seven years of unique data for training, focusing on optimization in the fields of translation and proofreading, supporting translation in Japanese, English, German and Simplified Chinese, and plans to expand support for more languages in the future.
According to DeepL's blind test results, the new model outperformed the traditional model in terms of translation quality. The test showed that the new model performed 1.7 times better in "Japanese-English translation" and "Simplified Chinese-English translation", and 1.4 times better in "German-English translation".
In addition, in comparisons with Google Translate, GPT-4, and Microsoft Translate, the new model was rated as "better translation quality" 1.3 times more often than Google Translate, 1.7 times more often than GPT-4, and 2.3 times more often than Microsoft Translate. To achieve the same translation quality as the new model, Google Translate requires twice as much editing work, while GPT-4 requires three times as much editing work.
Currently, the new generation of language models is available in the DeepL Pro plan, and users can switch to the new model with just a button. The monthly fee for the DeepL Pro plan starts at 1,150 yen (tax included) and is divided into three packages: Starter, Advanced, and Ultimate.