DeepL has achieved a major technological feat by translating the entire internet—more specifically, the Common Crawl dataset—from one language into any other within a span of just 18 days. This leap in performance was made possible through the integration of NVIDIA’s newest artificial intelligence chips, running on a high-performance supercomputer located in Sweden. According to DeepL, the system delivers translation speeds that are 30 times faster overall, with the full internet-scale translation itself being completed 11 times faster than their previous benchmarks.
The Common Crawl dataset is a vast public archive of web pages collected over time, frequently used for training and evaluating large-scale language models. By translating this comprehensive data set across multiple languages, DeepL has taken a significant step forward in multilingual AI capabilities. The translated content covers billions of web pages in dozens of languages, unlocking vast potential for global communication, training new AI systems, and enhancing translation quality across domains.
This upgrade not only accelerates the translation process but also enhances how DeepL tests and deploys its translation tools. DeepL’s system now supports any-to-any language translation—whether it’s English to Chinese or Arabic to Dutch—at a speed and scale previously unattainable. The system has become capable of handling all language directions without compromising accuracy or fluency, making it more versatile for enterprises, researchers, and developers around the world.
To demonstrate the sheer power of this system, DeepL presented examples showing how efficiently it handles large texts. For instance, translating the entire Oxford Dictionary now takes only two seconds. Translating an entire literary work, such as a full novel by Marcel Proust, completes in under a second. These figures illustrate the system’s extraordinary ability to process and understand complex linguistic structures rapidly.
Beyond speed, this accomplishment fundamentally alters DeepL’s internal development cycles. With faster data runs, the company can now iterate, experiment, and launch new translation tools more efficiently. This opens opportunities for businesses and institutions relying on real-time and large-scale translation, making content accessibility more inclusive and immediate across language barriers.
The achievement is not limited to just faster output—it’s about scaling translation in a way that can serve global demand in education, commerce, healthcare, diplomacy, and AI development. The integration of NVIDIA’s AI chips into this supercomputing setup provides DeepL with an infrastructure edge, allowing sustained progress in AI translation that keeps pace with the rapidly growing volume of global digital content.
