GLiNER-PII is inspired by the Gretel GLiNER PII/PHI models. Built on the GLiNER large-v2.1 base, it detects and classifies a broad range of Personally Identifiable Information (PII) and Protected Health Information (PHI) in structured and unstructured text. It is non-generative and produces span-level entity annotations with confidence scores across 55+ categories. This model was developed by NVIDIA.
This model is ready for commercial/non-commercial use.
This model was developed based on urchade/gliner_large-v2.1 Number of model parameters: 5.7 × 10^8
Input:
Input Type(s): Text Input Format: UTF-8 string(s) Input Parameters: One-Dimensional (1D) Other Properties Related to Input: supports structured and unstructured text
Output:
Output Type(s): Text Output Format: String Output Parameters: One-Dimensional (1D) Other Properties Related to Output: List of dictionaries with keys {text, label, start, end, score}
The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.
Model Version(s):
nvidia/gliner-pii
Version: v1.0
Training and Evaluation Datasets:
Training Dataset
Link:nvidia/nemotron-pii Data Modality: Text Text Training Data Size: ~100k records (~10^5, <1B tokens) Data Collection Method: Synthetic Labeling Method: Synthetic
Properties:
Synthetic persona-grounded dataset generated with NVIDIA NeMo Data Designer, spanning 50+ industries and 55+ entity types (U.S. and international formats). Includes both structured and unstructured records. Labels automatically injected during generation.
Data Collection Method: Hybrid: Automated, Human Labeling Method: Hybrid: Automated, Human
Evaluation Results
From the combined evaluation across Argilla, AI4Privacy, and Gretel PII datasets:
Benchmark
Strict F1
Argilla PII
0.70
AI4Privacy
0.64
nvidia/Nemotron-PII
0.87
We evaluated the model using threshold=0.3.
Inference:
Acceleration Engine: PyTorch (via Hugging Face Transformers) Test Hardware: NVIDIA A100 (Ampere, PCIe/SXM)
Usage Recommendation
First, make sure you have the gliner library installed:
pip install gliner
Now, let's try to find an email, SSN, and phone number in a messy block of text.
from gliner import GLiNER
# 1. Define our new text
text = "Hi support, I can't log in! My account username is 'johndoe88'. Every time I try, it says 'invalid credentials'. Please reset my password. You can reach me at (555) 123-4567 or johnd@example.com"
# 2. Define the labels we're hunting for.
labels = ["email", "phone_number", "user_name"]
# 3. Load the PII model
model = GLiNER.from_pretrained("nvidia/gliner-pii")
# 4. Run the prediction at given threshold
entities = model.predict_entities(text, labels, threshold=0.5)
NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.
For more detailed information on ethical considerations for this model, please see the Bias, Explainability, Safety & Security, and Privacy Subcards.
Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns here.
Source Model Repo
This model is derived from nvidia/gliner-PII. See the upstream repository for the original safetensors weights, training data, and the full upstream model card.
ONNX Weights
ONNX weights added by SAS — converted from the upstream safetensors checkpoint.
File in this repo: model.onnx.
Using this Model with the SAS GLiText API
This repo is consumed by the SAS GLiText product. To download it onto a SAS GLiText server:
