Medmorf

LLM only (recommended on WebGPU devices): Qwen3 handles all PII detection in a single pass. Best accuracy, especially with the 4B or 8B model.

NER + LLM: Run a NER model first, then let the LLM verify and catch remaining PII. Useful for additional coverage or comparison.

NER only: Run just the selected NER model. Fastest option, no GPU required, but lower accuracy.

Hardware: LLM modes need a modern GPU or Apple Silicon Mac with WebGPU (Chrome/Edge 113+, Safari 18+). The 4B model needs ~3.4 GB GPU memory; pick a smaller model on devices with limited VRAM. NER-only mode runs on CPU.

• Qwen3 0.6B — ~1.4 GB VRAM. Smallest & fastest LLM. Good for quick scans.
• Qwen3 1.7B (preferred default when feasible) — ~2 GB VRAM. Good balance of speed and quality.
• Qwen3 4B — ~3.4 GB VRAM. Higher quality, needs more headroom.
• Qwen3 8B — ~5.7 GB VRAM. Highest quality, needs a powerful GPU.
• Multilingual PII NER — ~280 MB. XLM-RoBERTa, no GPU needed. Names, addresses, dates, IDs in 8+ languages.
• GLiNER PII Edge — ~46 MB. Zero-shot, no GPU needed. Best for English.
• Multilingual BERT NER — ~100 MB. Lighter general-purpose NER (people, places, organizations).
• OpenAI Privacy Filter — ~1.5 B params (q4 ≈ 800 MB). OpenAI's bidirectional token classifier with 8 PII categories: person, email, phone, address, date, URL, account number, secret. Runs in-browser via Transformers.js + WebGPU; WASM/CPU is not supported for its quantized embedding op in this browser stack. Primarily English with multilingual robustness reported. Pick it under Advanced settings → NER Model with the NER only or NER + LLM pipeline.

Always manually review the output. First load downloads the selected model; subsequent runs use the browser cache.