Pangeanic

What Is Pangeanic?

A regional tax office needs 25,000 civil servants to read documents written in Spanish, Catalan, Galician, and Basque without waiting days for a human translator, and without sending taxpayer records to a US cloud API. That is roughly the brief Spain's Agencia Tributaria handed to Pangeanic in 2020, and it explains more about the company than any feature list.

Pangeanic is a Valencia-based language technology company that builds adaptive machine translation engines, PII anonymization tools, and data-for-AI pipelines for organizations that cannot use an off-the-shelf translation API. Manuel Herranz founded the business in 2000 as BI Europa, the European arm of Japan's B.I Corporation, then led a management buyout in 2005 that rebranded it Pangeanic under independent, founder-led ownership. In 2010 the company launched PangeaMT, a commercial adaptive engine built on the open-source Moses statistical MT platform — early enough that Pangeanic predates most of the neural MT vendors it now competes against.

Two decades on, the product line has widened well past translation. Masker handles PII detection and anonymization. A data-for-AI division builds synthetic and human-annotated datasets for LLM training. Machine Translation Quality Estimation (MTQE) scores output without a human reviewer in the loop. All of it runs through the ECO platform, deployable as SaaS, on-premises, or fully air-gapped for clients who cannot risk data leaving their own infrastructure. That's a requirement that shows up often in government and defense-adjacent procurement.

Key Features

How does PangeaMT's adaptive engine differ from a generic translation API?

Deep Adaptive AI Translation (DAAIT) is Pangeanic's term for engines that retrain on human corrections as translators work, rather than shipping a fixed model and hoping it fits your domain. A generic engine translates "resolución" the same way in a legal filing and a product manual; an adapted engine learns which meaning your organization actually uses and applies it consistently across every subsequent document.

This matters most for specialized vocabulary — legal, medical, technical, or government terminology that a general-purpose model was never trained on with any depth. Pangeanic quotes BLEU and COMET scores alongside human expert evaluation to measure the gap between a generic engine and an adapted one, and 200+ language pairs are supported across the platform.

What does Masker actually redact, and how?

Masker scans documents for personally identifiable information — names, addresses, ID numbers, financial details. Available techniques include tagging (replacing "John Smith" with "NAME1"), substitution (swapping in a realistic but fake replacement), gaps, and full redaction. It's built to run inside the same pipeline as translation, so a document can be anonymized and translated in a single pass rather than two separate vendor relationships.

For sectors like insurance, legal services, and healthcare, that combination is the actual selling point: translating patient records or claims documents without exposing personal data to whoever handles the translation step.

What is the data-for-AI service, and who needs it?

Enterprises training or fine-tuning language models increasingly need multilingual, domain-specific, or synthetic datasets that don't exist off the shelf. Pangeanic's data-for-AI division builds these datasets — combining anonymized real data, synthetic augmentation, and human annotation — aimed at organizations that need training data without exposing production records to a third-party model provider. It's a natural extension of two decades spent building translation memories and terminology databases, which are themselves a form of structured multilingual data.

Where does Pangeanic run — cloud, on-prem, or both?

All three. The ECO platform is available as multi-tenant SaaS for standard workloads, but Pangeanic also supports on-premises and fully air-gapped deployment for clients who cannot allow data to leave their own network. That's a real requirement for defense, law enforcement, and some government agencies. Veritone deployed an on-premises MT engine farm for law enforcement transcript processing through Pangeanic and reported 90% annual cost savings versus outsourced human translation. Small Language Models are also available for lightweight or resource-constrained environments where a full neural engine isn't practical.

How does CAT tool and API integration work?

Pangeanic exposes a REST API and integrates with SDL Trados, memoQ, and Memsource, plus TMX import/export for translation memory portability and CMS plugins for direct publishing workflows. It's a smaller integration list than DeepL's, which reaches into Microsoft Office, Google Workspace, and Crowdin, but it covers the tools professional translation teams already standardize on.

Pricing

Pangeanic doesn't publish a full price list, and that's deliberate — enterprise MT vendors selling adaptive, domain-trained engines rarely do, because the actual cost depends on volume, customization depth, and deployment model. What is publicly visible: a "PangeaMT Standard" plan reportedly runs around €1,000 for a 12-month term (roughly €100/month), and a shorter 3-month "Silver" plan covers up to 1,000 pages for around €500. Treat both figures as directional rather than a locked-in quote — we could not verify them against a current, dated published price list, which is common for custom-quote products at this tier.

Everything past that entry point — adaptive domain engines, on-premises or air-gapped deployment, Masker anonymization, and data-for-AI services — is priced individually. Pangeanic states a 24-hour turnaround on custom quote requests, which is fast for enterprise sales cycles. Budgeting still requires a conversation, though, before you can compare costs against DeepL's published API rates or Google Cloud Translation's per-character pricing.

There's no ongoing free tier. A free, non-binding demo covers up to 50 pages of translation or anonymization. Pangeanic has also offered a two-week trial specifically to compare a generic engine's output against an adapted one on your own sample content. That's useful given the whole pitch is "our engine learns your terminology" — it's exactly what the trial is built to prove.

EU Compliance & Privacy

Spain is a full EU member state, which puts Pangeanic squarely under GDPR by default rather than relying on adequacy decisions or standard contractual clauses to bridge a non-EU jurisdiction. The company states its platform runs zero-retention data policies and processes data within the EU.

Certification-wise, Pangeanic holds ISO 27001 for information security — notable because it was reportedly the first Spanish translation and NLP company to achieve it. It also holds ISO 9001 for quality management (held since 2007), ISO 17100:2015 for translation services, and ISO 13485:2016, a standard specific to medical device documentation. That last one signals real investment in regulated-sector compliance rather than a general security badge bolted on for sales collateral.

The government client list reinforces the same point from a different angle. A public-sector procurement process for a national tax authority or a national parliament involves security vetting that a typical SaaS sales cycle doesn't. Agencia Tributaria has used Pangeanic's ECO platform since 2020 to translate documents for around 25,000 staff across Spain's co-official regional languages. The Congress of Deputies separately contracted Pangeanic for AI-assisted transcription of parliamentary speeches delivered in Catalan, Galician, and Basque — building on transcription work Pangeanic has done for the Valencian regional parliament since 2021. Pangeanic has also participated in EU-funded research, including the Connecting Europe Facility-funded IADAATPA consortium project and a translation R&D project co-financed through Spain's CDTI and the EU's ERDF Multi-Regional Programme.

Who It's Best For

Regulated enterprises and public-sector bodies that need translation or anonymization pipelines where data cannot leave EU (or their own) infrastructure. On-premises and air-gapped deployment options aren't something DeepL or Google Translate offer at all.

Organizations with genuinely specialized vocabulary — legal, medical, technical, financial — where a generic MT model produces translations that are fluent but wrong in the details that matter. The adaptive engine model exists specifically to close that gap over time.

Teams needing more than translation alone, particularly anonymization plus translation in a single pipeline, or multilingual training data for their own AI models. Bundling those services with an established MT vendor beats stitching together three separate contracts.

If you just need fast, good-quality translation for everyday content — emails, marketing copy, general business documents — DeepL or Google Translate will get you there faster, cheaper, and with a far smoother self-serve experience. Pangeanic's value shows up specifically at the point where "generic and fast" stops being good enough.

The Verdict

Pangeanic earns its EU-compliance credentials the hard way: through government procurement, ISO 27001 certification, and participation in EU-funded research rather than marketing copy alone. The adaptive engine approach and the anonymization-plus-translation combination are genuinely differentiated, and the on-premises/air-gapped deployment options solve a real problem for regulated buyers that DeepL and Google Translate simply don't address.

Transparency is the trade-off here. A single named entry-level price point doesn't make up for the fact that most real deployments require a sales conversation. Nor can a company of roughly 120 people match DeepL's or Google's polish for casual, self-serve use. For an enterprise or public-sector buyer with specialized terminology and strict data-handling requirements, that trade-off is worth making. Someone who just wants to translate an email won't find that trade-off worth it — go use DeepL instead.

Frequently Asked Questions

Is Pangeanic GDPR compliant, and where is data hosted?

Yes. Pangeanic is headquartered in Valencia, Spain, an EU member state, and states its platform runs GDPR-compliant, zero-retention data policies. The company holds ISO 27001 information security certification — the first Spanish translation and NLP company to do so — plus ISO 9001. Data processed through its ECO platform is hosted in the EU.

How does Pangeanic compare to DeepL?

DeepL is a self-serve translation product with a generous free tier and consumer-friendly apps, strongest on general-purpose text quality across 30+ languages. Pangeanic targets enterprise and government buyers needing adaptive, domain-trained engines, on-premises or air-gapped deployment, and adjacent services like PII anonymization and data-for-AI — capabilities DeepL does not offer. For everyday translation, DeepL is faster to start with; for regulated or highly specialized workloads, Pangeanic's customization goes deeper.

Does Pangeanic have a free tier?

No ongoing free tier. Pangeanic offers a free, non-binding demo covering up to 50 pages of translation or anonymization, plus a reported 2-week trial for testing adaptive engine accuracy against a generic model. Beyond that, PangeaMT Standard is a paid entry plan, with everything above it quote-based.

How much does Pangeanic cost?

Pangeanic's core pricing is custom and quote-based, typical of enterprise MT and data vendors. One publicly reported figure is PangeaMT Standard at roughly €1,000 per year (about €100/month), with a smaller 3-month "Silver" plan around €500 for up to 1,000 pages. Adaptive domain engines, on-premises deployment, anonymization, and data-for-AI services are priced individually — Pangeanic states a 24-hour turnaround on custom quote requests.

Which government agencies use Pangeanic?

Pangeanic has been a contractor for Spain's Agencia Tributaria (Tax Agency) since 2020, translating documents for roughly 25,000 civil servants across Spain's co-official languages via its ECO SaaS platform. The Congress of Deputies also awarded Pangeanic a contract for AI-assisted transcription of parliamentary speeches in Catalan, Galician, and Basque, building on transcription work for the Valencian regional parliament since 2021.