ModernMT
Discontinued Last verified: September 2026Adaptive neural machine translation that learns from your corrections in real time, built for professional translators and LSPs. Discontinued: ModernMT is evolving into Lara and will sunset by the end of 2026.
What is ModernMT?
ModernMT is an adaptive neural machine translation system developed by Translated, the Rome-based language services company behind Matecat and T-LM. Unlike static engines such as Google Translate or the consumer tier of DeepL, ModernMT learns in real time from each correction a translator makes, a feature called instance-based adaptation. Every post-edit becomes training data for the very next sentence, so if you correct a product name, preferred phrasing, or industry term once, ModernMT uses that context going forward without needing a full model retrain. Translated's own comparison table lists 200+ supported languages, and the system was sold primarily as an API product with both self-hosted enterprise deployments and a cloud API through the ModernMT Hub. Translated positions it against DeepL and Google for agencies and in-house localization teams that need context-sensitive quality. ModernMT integrates directly with CAT tools like memoQ, Trados, Phrase, and Matecat, and is widely used by language service providers handling high-volume multilingual content. Status update, checked 2 August 2026: ModernMT is being retired. Translated has folded ModernMT into a successor product called Lara, and modernmt.com now states that ModernMT is evolving into Lara and will sunset by the end of 2026. New registrations on modernmt.com are closed and prospective users are sent to laratranslate.com instead; existing ModernMT API keys are stated to keep working until 31 December 2026, after which integrations must move to a Lara key. Translated's own ModernMT-versus-Lara comparison table now lists 200+ supported languages for ModernMT. The last pricing ToolChase was able to verify, in May 2026, was API access at roughly $15-20 per million characters depending on language pair and volume, a self-hosted personal license from around €40-60/month, and enterprise plans on request; modernmt.com no longer publishes any pricing page, so treat those figures as historical rather than current and expect a Lara quote instead. It was never a consumer web app, and there is no free public translator page, which kept the product firmly focused on professional workflows.
ModernMT demo video
Watch ModernMT's official demo to see ModernMT in action before reading our full review.
Official video by ModernMT via YouTube, embedded for reference. ToolChase does not host or claim this video.
⚡ Quick Verdict
Professional translators, LSPs, and in-house localization teams who want MT that learns from corrections in real time
Casual users or anyone who just needs a quick web translator for a paragraph of text
Closed to new signups as of August 2026 · last verified May 2026: API from ~$15 per 1M characters · Enterprise custom quote · Personal license from ~€40/mo
No free consumer tier, and new signups are closed as of August 2026
Real-time adaptive learning from every post-edit without retraining the base model
Being retired: sunsets by the end of 2026 and is closed to new registrations
Bottom line: ModernMT scored 4.3/5 as a CAT-tool engine where context adaptation materially improved productivity, but it is being retired: it sunsets by the end of 2026 and no longer accepts new signups. New buyers should evaluate its successor, Lara, or an alternative such as DeepL Pro.
Pricing
Historical, not currently purchasable. Checked 2 August 2026: modernmt.com no longer publishes a pricing page and is closed to new registrations. The site now states that ModernMT is evolving into Lara and will sunset by the end of 2026, so none of the plans below can be bought. The figures are the last ones ToolChase verified, in May 2026, and are retained for existing customers and for reference. We have not substituted Lara pricing, because that is quoted per account.
API, from ~$15 per million characters: Pay-per-character API access to the ModernMT Hub with automatic instance adaptation. Language pair and volume affect the final rate.
Personal License, from ~€40-60/month: Self-hosted personal ModernMT for individual professional translators, typically bundled with Matecat.
Enterprise, custom quote: On-premise deployment with dedicated model training, private corpora, custom language pairs, and integration consulting. Usually priced per core or per million words with annual commitments.
Key Features
- Instance-based adaptation, learns from every correction in real time
- 200+ languages per Translated's own ModernMT versus Lara comparison
- Direct integration with Matecat, memoQ, Trados, Phrase, Across
- REST API with OpenAPI spec and multi-region endpoints
- On-premise deployment for data-sensitive enterprises
- Context window that carries terminology across a full document
- Translation memory integration and fuzzy-match suggestions
- Domain adaptation from customer-provided corpora
- Custom glossary enforcement
Pros & Cons
Pros
- Genuinely adaptive, corrections improve the next sentence, not next quarter
- Strong integrations with professional CAT tools
- On-premise option for regulated industries
- API pricing was transparent, with enterprise volume discounts
Cons
- No free consumer web translator, API only
- Less well-known than DeepL or Google outside the localization industry
- Personal license pricing can add up for full-time freelancers
FAQ
What makes ModernMT adaptive?
ModernMT uses instance-based adaptation, when you post-edit a translation in your CAT tool, the correction is immediately available as context for the next segment without retraining the underlying model. Traditional MT engines require periodic full retraining to absorb corrections, which can take days or weeks. ModernMT's approach means if you correct gadget to widget in segment 1, segment 2 of the same document will already know your preference. This is particularly valuable for long documents with consistent terminology.
Is ModernMT better than DeepL for professional translators?
For single-sentence quality on European languages, DeepL still wins in most blind evaluations. But for full-document workflows where the translator is post-editing at scale, ModernMT often produces higher end-of-day productivity because each correction reduces downstream editing. LSPs that have benchmarked both tend to use DeepL as a fallback and ModernMT as the primary engine inside CAT tools.
Does ModernMT have a free tier?
No, and as of 2 August 2026 you can no longer sign up at all: modernmt.com states that new registrations on ModernMT.com are closed and directs prospective users to laratranslate.com/sign-up. ModernMT previously offered a free API trial with a limited character quota, but there was never an ongoing free consumer product. If you want a free web translator, Google Translate or the free tier of DeepL are the right choices.
What languages does ModernMT support?
Translated's own ModernMT versus Lara comparison table, published on modernmt.com and checked 2 August 2026, lists 200+ supported languages for ModernMT, including all major European languages, Chinese, Japanese, Korean, Arabic, Hindi, and several low-resource pairs. Translated no longer publishes a language-pair count. Bear in mind that ModernMT is scheduled to sunset by the end of 2026, so new language coverage now lands in Lara rather than ModernMT.
Can I run ModernMT on-premise?
Not as a new customer. modernmt.com is closed to new registrations and states that ModernMT is evolving into Lara and will sunset by the end of 2026. On-premise deployment was a core enterprise offering: ModernMT shipped a Docker-based installer that ran on Linux servers with GPU acceleration. This was the typical deployment for law firms, pharma, and financial services who could not send content to any external API. Pricing was custom and usually included setup, custom corpus training, and a dedicated support contract.
How does ModernMT compare to SYSTRAN?
Both target enterprise MT, but they differ in philosophy. SYSTRAN offers broader language coverage and a longer track record in government and defense. ModernMT focuses on adaptive quality inside professional CAT workflows. For a localization agency, ModernMT often feels more productive. For a government agency translating rare language pairs, SYSTRAN is usually the safer pick.
Does ModernMT integrate with Phrase or memoQ?
Yes, direct integrations exist for Phrase (formerly Memsource), memoQ, Trados Studio, Matecat, and Across. Setup typically involves pasting an API key into the CAT tool's MT plugin settings. Once connected, ModernMT appears as a machine translation provider alongside DeepL, Google, and Microsoft, and post-edits feed back into the adaptive engine automatically. Note that ModernMT is evolving into Lara and will sunset by the end of 2026, so new integrations should be built against Lara instead.
📋 Good to know
No longer possible for new users: modernmt.com closed registrations ahead of the end-of-2026 sunset and points new signups to laratranslate.com. Existing customers keep using their API key in their CAT tool until 31 December 2026, then swap in a Lara plugin and key.
Cloud API retains data for the length of your session to support adaptation. On-premise deployment keeps all data inside your network. Translated is GDPR compliant.
Not applicable now that ModernMT is being retired. The relevant decision is when to migrate: Translated reviews your setup and moves your Translation Memories and glossaries to a Lara plan before the end-of-2026 sunset.
Low for CAT-tool users, plug in an API key and go. Medium for on-premise, expect 1-2 weeks of setup and corpus preparation.
