DeepL's next-gen LLM outperforms ChatGPT-4, Google, and Microsoft for translation quality

author_by DeepL Team

What you need to know about our latest language model:

  • Our next-generation (“next-gen”) language model outperforms Google Translate, ChatGPT-4, and Microsoft for translation quality
  • The new LLM's translations require fewer edits, with Google needing 2x and ChatGPT-4 needing 3x more edits to achieve the same quality
  • Built using our own groundbreaking, specialized LLM technology and proprietary training data, designed specifically for translation
  • The same enterprise-level security you’re used to for Pro customers

As part of our mission to provide the highest quality translations to businesses around the world, we’re excited to introduce our next-generation language model to DeepL Translator. 

Powered by Language AI, alongside bespoke translation and writing technology, our new language model takes translation excellence to the next level—requiring fewer edits and improving knowledge worker productivity. This helps your enterprise save more time and money translating by further streamlining global communication.

Read on to learn more about our latest language model—then be sure to enable the new model (starting with Japanese, German, and Simplified Chinese to and from English) via our in-product selector, and give it a try for yourself.

What can you expect from our next-gen model?

Images of two people using large language models and natural language processing for translation

There’s lots to be excited about with our next-gen model. Let’s cover the most important topics.

Groundbreaking technology that outperforms competitors 

DeepL’s models are trained on the highest-quality linguistic data, with an unwavering focus on excellence. Combining world-class AI with the expertise of thousands of hand-picked language specialists who “tutor” the model, DeepL consistently provides best-in-class translation.

In fact, recent blind tests show that language experts routinely prefer DeepL's translations. According to the data, DeepL’s translation is preferred:

  • 1.3x more often than Google Translate
  • 1.7x more often than ChatGPT-4
  • 2.3x more often than Microsoft

But that’s not all. Thanks to specialized LLMs, which are uniquely tuned for language, DeepL’s more human-like translations also have a reduced risk of hallucinations and misinformation. This is especially important for enterprises looking to scale operations with a trusted AI solution.

Furthermore, unlike general purpose models that train with data pulled from the public internet, DeepL’s model leverages over seven years of proprietary data specifically tuned for translation and content creation.

Highest-quality translations on the market

Starting with Japanese, German, and Simplified Chinese to and from English, with many more languages to come, this step forward continues to improve on the top-notch translations that DeepL’s web translator provides. With the innovations of the next-gen model, DeepL Translator provides the highest quality translations on the market.

The quality of our next-gen models was confirmed via our previously mentioned blind tests, which were conducted with leading language experts. In these tests, the quality of the LLM-powered translations compared to the “classic” model was determined—with the new model coming out on top for each test.

  • Japanese ←→ English: the quality of the next-gen model is 1.7x better than the classic model
  • Simplified Chinese ←→ English: the quality of the next-gen model is 1.7x better than the classic model
  • German ←→ English: the quality of the next-gen model is 1.4x better than the classic model

This improved translation quality will not only save your company time, but also allow your teams to streamline communications across global markets.

Improve productivity to save more time and money

Our next-generation model will help your enterprise get even more value from DeepL. In fact, blind tests also found that DeepL’s translations require fewer edits than competitors, with Google Translate needing 2x more edits and ChatGPT-4 needing 3x more edits to achieve the same quality. 

Thus, our new LLMs improve knowledge worker productivity by decreasing the time spent on translation and copy editing, allowing your teams to focus on the jobs that matter the most.

In fact, in the 2024 study conducted by Forrester Consulting, “The Total Economic Impact™ of DeepL”, it was discovered that DeepL reduced translation time by 90% and cut translation workloads by 50%—resulting in a 345% ROI.* 

The same security and privacy you’re used to

Just like before, our next-gen model is built with security at its core thanks to:

  • Our proprietary data centers
  • The highest level of certification and compliance standards:
    • ISO 27001 certification
    • GDPR compliance
    • SOC 2 type 2 report
  • A commitment to never using Pro customer data to train our models

Take your enterprise communication to the next level with higher-quality translations

Illustration of a globe showing human language and large language models with floating characters

Our next-gen model represents another major step forward in providing your enterprise with increasingly powerful features. With this significant increase in translation quality—which improves upon our “classic” model and outperforms competitors—we’re raising the bar for AI translation tools everywhere.

How can your company start making use of our next-gen model? It’s easy: Pro users can switch between the next-gen LLM and “classic” model with the click of a button.

Try out the next-gen model with your Pro plan, or consider upgrading to Pro.

*DeepL’s Forrester study was conducted in 2024 using a composite organization, which tested the “classic” model.