ModernMT

ModernMT is an adaptive neural machine translation platform developed by Translated, a Rome-based language services company with over two decades of experience in the translation industry. Unlike static translation engines that produce the same output regardless of context or user feedback, ModernMT was designed around a fundamental premise: translation quality improves dramatically when the system can learn continuously from professional translators and adapt to the specific domain, style, and terminology of each document being translated. The platform combines large-scale neural models trained on billions of segments with a real-time adaptation layer that incorporates corrections within seconds, without requiring retraining cycles, batch processing, or dedicated model fine-tuning. This architecture lets the engine behave like a custom-trained model for each user, project, and document, while keeping the simplicity of a general-purpose API.

The system supports over 200 languages and more than 40,000 language pairs, covering the vast majority of localization needs for global enterprises, e-commerce platforms, software vendors, and translation agencies. Its document-level context awareness means that ModernMT considers surrounding sentences, not only the single segment being translated, which results in more coherent pronoun resolution, tense consistency, and terminology usage across long documents. The engine integrates natively with the main CAT tools used by professional linguists, including Trados Studio, memoQ, Phrase, XTM, and Matecat, so translation memories and glossaries become additional adaptation signals rather than static reference files. For enterprises, ModernMT offers API access, a REST endpoint, a plugin ecosystem, and on-premises deployment options for customers with strict data-residency or confidentiality requirements.

A distinctive feature of ModernMT is its Human-in-the-Loop architecture: when professional translators post-edit machine output, their corrections feed back into a user-specific adaptation space that preserves privacy while progressively improving quality for that account. Translated also operates a global network of over 300,000 professional translators through its Matecat and Translated.com platforms, which can be combined with ModernMT for full hybrid workflows where AI produces the first pass and humans refine high-stakes content. In 2024 Translated publicly introduced Lara, a large language model-based translation system that sits alongside ModernMT and targets highest-quality contextual translation for complex documents, marketing copy, and creative content; Lara is marketed as closing the gap with professional human translators on a measurable edit-distance metric.

Typical customers include Airbnb, Uber, Amazon MGM Studios, and localization departments at software and media companies, where ModernMT handles millions of characters per day. Pricing is transparent and consumption-based, with a flat monthly plan for individual translators, a pay-per-million-characters Adaptive tier for developers integrating the API, and higher-tier Human-in-the-Loop and Localization Teams plans that bundle professional post-editing and dedicated support. ModernMT is a practical choice for teams that need better-than-generic machine translation, want to preserve their terminology and brand voice, and prefer an engine that measurably improves as their linguists use it, rather than a black-box commodity translation service.