PIIMB PII Masking - Avg F2: leaderboard

PIIMB evaluates zero-shot PII masking over OpenPII, Gretel, Nemotron-PII, and Privy datasets, with F2, F1, precision, recall, and NER metrics.

Metric: Average F2 (%). Source: huggingface.co. 35 models tracked.

Top models

#ModelScore
1OpenMed privacy-filter-multilingual-v291.68
2knowledgator gliner-stream-pii-v1.089.93
3OpenMed OpenMed-PII-SuperClinical-Large-434M-v188.52
4OpenMed privacy-filter-nemotron-v287.34
5OpenMed OpenMed-PII-SuperClinical-Small-44M-v186.79
6antoniogr7 pii-identifier-v0.186.23
7kalyan-ks ettin-68m-nemotron-pii84.11
8fastino gliner2-privacy-filter-PII-multi83.48
9kalyan-ks ettin-32m-nemotron-pii83.08
10nvidia gliner-PII82.98

Interactive version: theaggregate.ai/benchmark?slug=piimb-pii-masking-avg-f2 · How It Works · Data refreshed daily, snapshot 2026-09-05.