DeepL reveals next-gen language model for superior translations
DeepL has introduced its next-generation language model, leveraging a highly specialised LLM technology specifically designed for translation and editing.
This update is considered a significant advancement in DeepL's Language AI platform for businesses, setting a new industry standard for translation quality and performance.
The new solution integrates three unique capabilities. Firstly, it features a Specialised LLM that is uniquely tuned for language, resulting in more human-like translations and writing while reducing the risk of hallucinations and misinformation.
Secondly, unlike general-purpose models that train on public internet data, the DeepL model uses over seven years of proprietary data aimed at content creation and translation. Thirdly, the model benefits from human tutoring by thousands of language experts specially trained to deliver best-in-class translations.
Blind tests comparing the new DeepL model with its competitors highlight the improvement in translation quality. The DeepL translations were preferred by language experts 1.3 times more often than Google Translate, 1.7 times more often than ChatGPT-4, and 2.3 times more often than Microsoft.
Jarek Kutylowski, CEO and Founder of DeepL, commented, "We are a research-driven company to our core and are investing heavily in pushing the boundaries of translation and writing quality, efficiency, and adaptability. Our goal is to empower our customers worldwide with the industry-leading technology they need to thrive and scale globally without a language barrier standing in the way."
Stefan Mesken, VP of Research, noted the practical benefits of the new model. "When testing the new model against competitors and our classic model, professional translators noted that Google requires 2x more edits and ChatGPT requires 3x more edits to get to the same quality," he said. "For businesses who translate a large volume of documents, this truly impacts their efficiency and productivity, meaning team members can focus on tasks with high-business value."
Since its founding in 2017, DeepL has become a preferred Language AI provider for businesses globally. It offers state-of-the-art translation and writing solutions crucial for international business growth. These solutions address communication challenges across various domains, including internal operations and customer support. Unlike general-purpose AI systems, DeepL's models are specifically tuned for language, resulting in more accurate translations and a reduced risk of errors.
The new model offers substantial improvements. Blind tests conducted with leading linguists found a 1.7x improvement with the new LLM compared to the old model for translations involving English to Japanese and Simplified Chinese, and a 1.4x improvement for English to German translations. These enhancements aim to save business users valuable time on manual editing, thereby improving communication efficiency across teams and markets.
The next-gen model is currently available for DeepL Pro customers for translations in English, Japanese, German, and Simplified Chinese. Additional languages are expected to be added soon. The new model can be activated through the web translator by selecting the next-gen model option, giving users greater control over their translation experience. DeepL Pro users also benefit from enterprise-grade security and compliance standards, including ISO 27001 certification and GDPR/SOC 2 type 2 compliance. Importantly, no Pro translations are used to train the models.
This new model comes amid a transformative period for DeepL. Earlier this month, the company launched DeepL for Enterprise, a language technology catering to businesses seeking to integrate AI solutions. DeepL's rapid expansion is reflected in its growing customer network, which includes over 100,000 businesses, governments, and organisations worldwide, including 50% of the Fortune 500. Additionally, DeepL announced it had raised USD $300 million in new investment at a USD $2 billion valuation in May 2024, led by investment firm Index Ventures.