Back to browse
A 0.3B model that redacts PII in all 24 EU languages offline

A 0.3B model that redacts PII in all 24 EU languages offline

by mipo57·May 13, 2026·6 points·3 comments

AI Analysis

●●●BangerNiche GemSolve My Problem

Native multilingual training covers GDPR Article 9 categories others skip.

Strengths
  • Trained end-to-end on real EU language data, not English translations bolted on.
  • Quantized ONNX weights enable CPU-only inference for real-time RAG pipelines.
  • Detects biometric and genetic data categories that standard OSS models miss.
Weaknesses
  • Model card lacks benchmark comparisons against Presidio or Microsoft Presidio.
  • Apache 2.0 license is great, but enterprise support path is unclear.
Category
Target Audience

Compliance engineers and privacy officers in EU companies

Similar To

Microsoft Presidio · Amazon Comprehend · Google Cloud DLP

Similar Projects