ModernMT

ModernMT Overview

ModernMT is an adaptive neural MT engine that learns in real time from your corrections — built on research from the University of Edinburgh's MT group, commercially run by Translated.

How ModernMT Works

ModernMT's core differentiator is adaptation: every correction a translator makes immediately reshapes output for subsequent segments — no offline retraining cycle. It also conditions on the full document context and can blend your translation memories. This yields engine output that converges toward your style within a project.

Key Capabilities

• Real-time adaptive learning from translator corrections
• Context-aware: uses surrounding document content for disambiguation
• TM blending: combines your memory with the neural engine
• Available inside major CAT tools (Trados, MemoQ, Smartcat)
• Pay-as-you-go cloud service with no fixed fee

Best Use Cases

• CAT-tool users wanting MT that improves within each project
• Long documents where consistency of emerging terminology matters
• Teams already storing TMs who want them to actively improve MT
• Cost-sensitive teams (pure usage pricing, no subscription floor)

ModernMT Pricing

ModernMT is usage-based: roughly $10-15 per million characters depending on volume, with no monthly minimum. It is often bundled free inside partner CAT tools' subscriptions (e.g., some Smartcat and Translated plans include Big ModernMT).

Frequently Asked Questions

How does ModernMT learn from my corrections?

Every confirmed segment feeds back immediately: the engine adapts its output for the rest of the document within seconds — no model retraining delay. Your corrections stay yours (privacy-safe adaptation).

How much does ModernMT cost?

About $10-15 per million characters, usage-based with no minimum. It is included free in several CAT tool subscriptions — check whether your tool bundles it before paying directly.

ModernMT vs DeepL — which is better?

Out of the box for European pairs, DeepL is generally more fluent. ModernMT wins in a workflow with translator corrections (it adapts to you) and on usage pricing. Best test: run both on a real document and measure post-editing distance.

Verdict

ModernMT is the pragmatic adaptive engine: connect it to your CAT tool, correct its output, and watch it converge to your voice within a document — at usage pricing with no subscription. It lacks DeepL's out-of-the-box fluency for European pairs and the platform features of full TMSes, but as the MT layer in a human-in-the-loop pipeline, its real-time learning is genuinely differentiated.