Another breakthrough in AI translation quality: DeepL's 2020 blind test results

Key Takeaways
- DeepL launched a new translation system in 2020, delivering another major improvement in translation quality.
- In blind tests, professional translators selected DeepL's translations four times more often than any competing system.
- The new neural networks improved DeepL's ability to capture meaning and produce professional, natural phrasing.
- DeepL's services have since grown to more than 1 billion users worldwide.
We’re pleased to inform you that we’ve launched a completely new translation system that represents another quantum leap in translation quality. The neural networks we use are far superior to previous technologies.
You can test and use them free of charge at DeepL.
We already made headlines in 2017 when we released DeepL Translator. Even then, it was superior in quality to large, high-tech companies’ translation systems and could translate longer passages without errors.
Our services quickly became popular, and more than 1 billion people use our specialized AI tools.
Now, our AI researchers have succeeded in achieving another breakthrough in translation quality. This new system has improved all the criteria that make for a good translation.
The new neural networks can also represent the meaning of the translated sentences in the target language much more precisely. And at the same time, they can often find more professional formulations.
A new benchmark in translation quality
This has prompted us to conduct new blind tests. We translated 119 lengthy passages from a wide variety of subjects using DeepL Translator and some competing systems.
Then, we asked professional translators to evaluate these translations and choose the best translation—without knowing which system produced which translation.
The translators selected the translations from DeepL four times more often than those from any other system.
We assume companies such as Google or Microsoft also have very good data for training neural translation networks. On the other hand, we’ve achieved many innovations in the mathematics and methodology of neural networks.
With their help, we can extend our lead.
Our vision for the future
It’s a special moment for our entire team that now billions of people have the opportunity to use these improvements.
At DeepL, our vision is to break down language barriers worldwide and bring cultures closer together.
To achieve this, we’re planning to add many more languages. In the coming months and years, we want to integrate translation technology wherever people communicate with each other, read texts, write, or do business with each other.
This vision also motivates many AI researchers and developers to join DeepL. If you too would like to help us bring the world closer together, explore DeepL Careers!
How far DeepL has come
Since this announcement, DeepL has grown well beyond translation alone. Today, DeepL's Language AI suite includes a wealth of sophisticated tools:
- Translation across 100+ languages, up from the handful supported in 2020
- DeepL Voice for real-time spoken translation
- DeepL Write, an AI-powered writing assistant
- DeepL API for integrating translation directly into business workflows
- Desktop, mobile, and browser apps, plus integrations with tools like Microsoft Office and Google Workspace
Explore DeepL's full Language AI suite to see where that 2020 vision has led.
Experience DeepL's quality for yourself
Try DeepL Translator free, and see why professional translators still choose it first.
Contact Sales if you’d like to explore how our specialized Language AI ecosystem can power your global enterprise.