EPIC
Evaluate machine translation output across 500+ language pairs in real time — with zero custom setup or training data required.
Generic
Broad coverage
Specialized
Vertical & language pair
Custom
Built by TAUS from your data
← COVERAGE
SPECIFICITY →
A generic Quality Estimation (QE) model is an off-the-shelf AI model designed to evaluate machine translation (MT) or LLM output instantly, without needing a human reference translation.
It’s trained on broad, multi-domain data rather than one company’s content, so it works across most content from day one. The trade-off: it judges translations against general language norms, not your terminology.
7.5B
words of TAUS training data
500+
language pairs
Scores every segment from 0.0 to 1.0 in real time, so you can set automated quality gates in your TMS on day one. It’s the default model in the EPIC free trial and always available to all users.
Out-of-the-box support for over 500 language pairs, including high-resource languages and low-resource pairs.
Start scoring immediately without gathering historical Translation Memories (TMs) or other training data.
Automatically approve high-quality translations to bypass human post-editing, instantly lowering translation costs.
Generic QE models deliver the highest return on investment when applied to high-volume, lower-risk or general-domain content streams, where broad fluency and structural accuracy are the primary criteria.
EXAMPLE APPLICATIONS
Live support chats, help desk tickets, knowledge base articles
WHY GENERIC QE WORKS
Requires fast turnaround; standard conversational and technical phrasing.
EXAMPLE APPLICATIONS
Product reviews, community forum posts, social media comments
WHY GENERIC QE WORKS
High volume, fast degradation of value; doesn’t enforce strict glossary terms.
EXAMPLE APPLICATIONS
Enterprise emails, internal documentation, memo translations.
WHY GENERIC QE WORKS
Focuses on general readability and semantic fidelity rather than external brand voice.
Integrating the Generic QE model into your software stack takes only a few minutes via the TAUS EPIC REST API.
DEFAULT METRIC
TAUS QE v2.0 — applied automatically if no custom metric ID is passed.
LATENCY
Sub-second response times for real-time applications such as live chat or dynamic UI strings, alongside asynchronous batch processing for high-volume localization.
INPUT FORMATS
Plain text (JSON payload) or standard localization files (XLIFF, TMX, CSV).
// Example REST API request (POST /2.0/estimate/)
{"source_language": "en-US","target_language": "de-DE","source_text": "Please submit your support requestusing the online form.","target_text": "Bitte reichen Sie Ihre Support-Anfrageüber das Online-Formular ein."}
Start scoring your machine translation output in real time with the TAUS Generic QE Model.