In the January/February 2021 issue I wrote about the surprising quality and ease-of-use of the OPUS-MT and OPUS-CAT products.1 When I interviewed developer Tommi Nieminen, an experienced professional translator and MT researcher, about the OPUS tools, he said—much to my surprise—that according to his testing, his products (the locally installed MT and the connector to CAT tools) are as good as or better than any other trainable machine translation (MT) engine.2 Except one: ModernMT.

ModernMT is a cloud-based adaptive solution for neural machine translation (NMT), just like a number of other systems (including the above-mentioned OPUS tools). Interestingly, though, the adaptive part of the technology is fundamentally different than its competitors because there are actually no changes in the base engine happening at any time. Instead, the system uses a technology called “instance-based adaptive NMT.” This consists of the translation request first being sent to a translation memory (TM) layer (which can consist even of a relatively small TM as long as it’s highly tuned). With similar segments found in that TM layer, the NMT engine’s “hyperparameters” are adapted on-the-fly so that a more suitable suggestion is generated.

The base engine, which is hosted in a number of data centers, including in Rome and San Jose, is trained in 46 languages (see the list at the homepage on the large data sets collected by Translated, the Italian language services provider and tech developer of the massive MyMemory TM. (Translated also owns a majority of the shares in ModernMT.) And while the base engines are retrained once or twice a year, your data effectively sits in the middle and adapts the MT suggestions to your style and terminology.

The benefit is that you don’t ever need to actually train a specific MT engine, but you can instead use a large generic engine whose suggestions are specialized by having the query parameters adapted as the translation is happening.

When ModernMT was first released as a commercial product in early 2019 it had two major problems: it was outrageously priced and used a completely outdated privacy concept that used customers’ data for general training purposes.

After being nailed with criticism from many sides (as well as little customer support), they completely overhauled both their pricing and their privacy considerations. There’s no longer any cross-training, and any data that you upload to enhance your own MT will be strictly used by you alone. Kind of what you would expect from a paid product these days.

The pricing is also in line with other tools. You pay by the number of characters as a language services provider or translation buyer (between $8 and $50 per million characters, depending on whether you want to train the engine as you translate or queue documents for a batch translation) and a monthly fee if you’re a freelance translator ($25).

You can use ModernMT via its website, but as a professional translator you would be more likely to use it via their application programming interface through a translation environment tool. Presently, you can use it directly within Trados Studio with an app you can download from the RWS app store, with memoQ through an already integrated plugin, and of course with MateCat (“of course” because MateCat is also owned by aforementioned Translated). Presently, the ModernMT team is working on a Chrome extension as well, which should allow for the use of any browser-based translation environment and ModernMT.

There is a free 30-day-trial period you can access at And unlike with other products, it might actually make sense to try it out first because, again, there is literally no setup aside from connecting to the MT and choosing a well-suited TM to “sit in the middle.”

  1. Zetzsche, Jost. “(More) Advanced Human-Computer Interaction For Translators,” The ATA Chronicle (January/February 2021), 30,
  2. Zetzsche, Jost. “Any Artistic Work, Especially One on a Large Scale,” The Tool Box Journal, Issue 21-4-324 (April 2021),

Jost Zetzsche is chair of ATA’s Translation and Interpreting Resources Committee. He is the author of Characters with Character: 50 Ways to Rekindle Your Love Affair with Language.

Remember, if you have any ideas and/or suggestions regarding helpful resources or tools you would like to see featured, please e-mail Jost Zetzsche at

The ATA Chronicle © 2022 All rights reserved.