privacy-filter-nemotron needs about 2.2 GB of memory at its main file (a 1.6 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 2 quants, from 1.6 GB to 2.8 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 0.1 GB.
Fine‑grained PII token‑classification model
privacy-filter-nemotron is a 1.4B‑parameter MoE model fine‑tuned for 55 fine‑grained PII categories. It outputs BIOES tags for each token and decodes spans with a constrained Viterbi algorithm. The model is available in GGUF format for local inference.
Good at
- Fine‑grained PII detection
- BIOES token classification
- Local inference with GGUF
Limits
- No vision support
- English‑only categories
- Requires custom privacy‑filter.cpp runtime
Best for
- PII redaction pipelines
- Data masking tools
- Compliance checks
Who it’s for. Developers building privacy‑aware applications that need precise PII extraction. The model is suited for backend services that process text and require accurate entity spans without external APIs.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 1.6 GB | 2.2 GB | Fits | Get file → | |
| F16 | 2.8 GB | 3.4 GB | Fits | Get file → |
For 8 GB VRAM use the F16 quantized file (2.82 GB, ~3.4 GB memory at 8 k context). For 4 GB VRAM the unknown quant (1.64 GB, ~2.2 GB memory) is recommended. Memory estimates are approximate.
What GPU do you need for privacy-filter-nemotron?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is F16 (2.8 GB file, about 3.4 GB in use).
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about privacy-filter-nemotron
How much VRAM does privacy-filter-nemotron need?
About 2.2 GB at its main file with an 8K context window. That is the 1.6 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run privacy-filter-nemotron on an 8 GB GPU?
Yes, with F16: the 2.8 GB file needs about 3.4 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of privacy-filter-nemotron GGUF?
It depends on the quant: file is 1.6 GB, F16 is 2.8 GB.
What context length does privacy-filter-nemotron support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 0.0 GB at 8K and 0.1 GB at 32K.
Where can I download privacy-filter-nemotron GGUF?
The files are on Hugging Face in the LocalAI-io/privacy-filter-nemotron-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does privacy-filter-nemotron use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run privacy-filter-nemotron on 4 GB VRAM?
Yes, use the unknown quant file (1.64 GB). It requires about 2.2 GB of VRAM at 8 k context.
Does the model support multiple languages?
No, this model only covers 55 fine‑grained English PII categories.
What runtime is required?
Use privacy-filter.cpp or LocalAI; the model is not a chat model and needs the custom privacy‑filter.cpp runtime.