Benchmark your redaction models.

Compare local redaction quality against a stronger audited baseline. Scores use only the shared direct-identifier scope; model-only detections are separated for review.

Ready No hosted benchmark run yet

Documents

Upload or paste to begin

Best F1

--%

Baseline optional

Local speed

-- tok/s

OpenAI PF + custom

Fastest file

--s

Latency tracked per file

Baseline

Reference model Grok baseline xAI Use for comparison Modelgrok-4.20 NR ModeNon thinking ScopeDirect IDs

Offline redaction model

Selected files

Paste text or upload files

Local latency

--s

Average per file

Baseline latency

--s

Optional xAI pass

Total cost

Baseline provider only

Benchmark inputs Use the hosted API with the same cookie auth as this UI. No API key is exposed to the browser.

Upload, paste, or select a benchmark file to begin

Run one file or a batch. The progress stream shows extraction, local model timing, baseline timing, and final per-document score data.

PDF · DOCX · XLSX · PPTX · RTF · HTML · TXT · CSV · MD · JSON · ZIP

Selected files Drop files here or use Upload files.

Pasted document name Pasted text

Document summary

Latency, throughput, precision, recall, and F1 by file.

Document	Words	Tokens	Local latency	Local tok/s	Local items	Grok latency	Grok tok/s	Precision	Recall	F1	Missed	Cost

Document details

Open a row for redacted output and audit details.