Leaderboard
Find which detector best hides the sensitive data you care about.
Start with what you need to protect
Compare personal data, secrets, and the datasets where detectors struggle.
Performance from easy to extreme
See how much sensitive data each detector hides completely at every difficulty level. Higher is better.
| Detector | Easy | Medium | Hard | Extreme |
|---|---|---|---|---|
| 85.3 | 76.9 | 60.7 | 36.3 | |
| 94.2 | 83.2 | 64.4 | 26.0 | |
| 75.4 | 69.8 | 53.8 | 26.4 | |
| 80.1 | 63.0 | 62.7 | 22.2 | |
| 61.3 | 60.6 | 39.6 | 27.7 | |
| 67.9 | 61.9 | 46.0 | 33.8 | |
| 75.9 | 61.6 | 46.5 | 31.9 | |
| 85.4 | 67.7 | 59.0 | 10.9 | |
| 84.4 | 64.6 | 51.6 | 17.1 | |
| 69.7 | 61.5 | 64.3 | 23.9 | |
| 80.9 | 62.7 | 56.5 | 9.4 | |
| 68.3 | 61.5 | 65.4 | 13.0 | |
| 62.7 | 52.1 | 23.9 | 19.9 | |
| 66.7 | 52.6 | 43.9 | 19.6 | |
| 52.9 | 45.7 | 35.4 | 27.3 | |
| 65.6 | 50.8 | 33.6 | 18.7 | |
| 77.9 | 55.1 | 52.9 | 15.6 | |
| 76.7 | 53.7 | 50.9 | 10.6 | |
| 63.5 | 48.1 | 20.6 | 21.7 | |
| 46.4 | 43.4 | 29.0 | 18.9 | |
| 59.5 | 43.3 | 38.3 | 19.3 | |
| 33.0 | 35.2 | 26.1 | 18.7 | |
| 39.8 | 36.9 | 41.0 | 11.3 | |
| 63.0 | 49.8 | 27.0 | 18.2 | |
| 31.3 | 32.5 | 25.9 | 16.4 | |
| 37.7 | 34.0 | 31.3 | 41.4 | |
| 41.4 | 32.5 | 45.3 | 17.5 | |
| 22.8 | 20.5 | 16.1 | 5.6 | |
| 20.8 | 17.7 | 21.9 | 25.6 | |
| 22.0 | 30.4 | 36.7 | 1.7 | |
| 16.5 | 15.8 | 18.9 | 8.8 | |
pplx+cpu-int8Not all datasets | 98.8 | 94.3 | 64.6 | — |
pplx+homoglyphNot all datasets | — | 93.8 | 77.2 | — |
gliner2-fastino-ruNot all datasets | 78.6 | 77.0 | 78.6 | 35.8 |
gliner-nvidia+sent300Not all datasets | 72.3 | 72.7 | 61.3 | — |
HiveTrace OmniNot all datasets | 74.8 | 71.3 | 46.9 | 36.3 |
gliner25-fastino+sent300Not all datasets | 74.6 | 72.5 | 56.5 | — |
gliner2-hivetrace-omni-ruNot all datasets | 79.0 | 70.5 | 62.8 | 35.9 |
nym-base+homoglyphNot all datasets | — | 71.3 | 67.6 | — |
openai-base-onnxNot all datasets | 76.7 | 71.7 | 46.6 | — |
gliner25-fastino+ov100Not all datasets | 75.6 | 71.9 | 58.4 | — |
nym-base+sent300Not all datasets | 74.3 | 66.3 | 57.9 | — |
pplx+sent300Not all datasets | 85.1 | 70.3 | 52.9 | — |
gliner25-fastino-ruNot all datasets | 75.4 | 64.2 | 62.5 | 31.4 |
nym-baseNot all datasets | 79.7 | 63.0 | 60.8 | 16.5 |
gliner-nvidiaNot all datasets | 82.1 | 67.3 | 51.7 | 25.9 |
nym-base+ov100Not all datasets | 74.1 | 63.6 | 59.8 | — |
ru-legal-ner+cpuNot all datasets | — | 62.9 | — | — |
nym-smallNot all datasets | 78.9 | 60.7 | 57.8 | 11.8 |
pplx+ov100Not all datasets | 86.9 | 68.2 | 52.1 | — |
gliner-nvidia+homoglyphNot all datasets | — | 68.7 | 45.0 | — |
nym-base+cpu-int8Not all datasets | 94.4 | 63.4 | 55.5 | — |
ru-legal-ner+homoglyphNot all datasets | — | 68.6 | 49.3 | — |
gliner-nvidia+ov100Not all datasets | 71.0 | 60.0 | 58.0 | — |
openmed-multilingualNot all datasets | 72.9 | 56.4 | 47.4 | 6.1 |
opf-ru-v2+homoglyphNot all datasets | — | 68.7 | 57.4 | — |
gliner-nvidia-ruNot all datasets | 67.5 | 59.2 | 58.9 | 21.4 |
gliner2-hivetrace-uniNot all datasets | 49.2 | 46.2 | 16.7 | 23.2 |
opf-ruNot all datasets | 75.0 | 58.2 | 55.2 | 11.7 |
openmed-nemotronNot all datasets | 65.2 | 49.1 | 43.3 | 8.3 |
presidio-ruNot all datasets | — | 54.5 | — | — |
gliner25-fastino+nochunkNot all datasets | 73.4 | 58.1 | 42.6 | — |
gliner25-fastino-ru+nochunkNot all datasets | 59.3 | 53.8 | 50.7 | — |
gliner-urchade-ruNot all datasets | 50.7 | 48.3 | 45.7 | 20.3 |
gliner2-vladlinv+homoglyphNot all datasets | — | 60.8 | 23.3 | — |
ru-legal-ner+sent300Not all datasets | 64.5 | 40.6 | 36.0 | — |
gravitee-small+cpu-int8Not all datasets | 71.1 | 51.9 | 25.8 | — |
kalyan-ettinNot all datasets | 65.0 | 45.3 | 40.4 | 8.8 |
gliner-multi-v21-ruNot all datasets | 51.8 | 40.2 | 50.3 | 25.0 |
gliner2-vladlinv-ruNot all datasets | 72.6 | 51.5 | 39.7 | 21.7 |
opf-ru-v2+sent300Not all datasets | 77.5 | 41.8 | 45.2 | — |
ru-legal-ner+ov100Not all datasets | 63.8 | 34.4 | 35.4 | — |
opf-ru-v2+ov100Not all datasets | 77.5 | 39.4 | 44.4 | — |
ru-legal-ner+cpu-int8Not all datasets | 43.4 | 44.4 | 29.8 | — |
credsweeper-nomlNot all datasets | 82.2 | 27.7 | 60.1 | — |
credsweeperNot all datasets | 76.4 | 0.5 | 55.9 | — |
fef2-secret-ru+cpu-int8Not all datasets | 3.6 | 31.9 | 16.5 | — |
davlan-mbert+cpu-int8Not all datasets | 55.3 | 25.0 | 27.2 | — |
ner-ru-gherman+homoglyphNot all datasets | — | 1.4 | 15.6 | — |
mmbert32k+cpu-int8Not all datasets | 50.9 | 29.1 | 30.9 | — |
gliner2-hivetrace-uni-ruNot all datasets | 26.3 | 21.5 | 31.3 | 16.1 |
mmbert32k+nochunkNot all datasets | 42.7 | 31.4 | 25.4 | — |
kalyan-ettin+cpu-int8Not all datasets | 41.6 | 21.0 | 30.5 | — |
ner-ru-gherman-onnxNot all datasets | 55.0 | 18.0 | 7.2 | — |
davlan-xlmr+cpu-int8Not all datasets | 51.4 | 18.7 | 18.1 | — |
ner-ru-gherman+cpu-int8Not all datasets | 52.1 | 16.9 | 6.9 | — |
detect-secretsNot all datasets | 45.7 | 17.1 | 24.3 | — |
deepsecretsNot all datasets | 40.9 | 14.2 | 34.7 | — |
spacy-alrosaitNot all datasets | 9.8 | 10.1 | 8.7 | 7.0 |
gliner-pii-edge+cpu-int8Not all datasets | 17.5 | 15.8 | 7.2 | — |
trufflehogNot all datasets | 23.6 | 0.1 | 12.5 | — |
noseyparkerNot all datasets | 16.2 | 1.0 | 23.3 | — |
titusNot all datasets | — | 1.0 | 24.3 | — |
kingfisherNot all datasets | — | 1.0 | 13.9 | — |
gliner-pii-base+cpu-int8Not all datasets | 0.0 | 1.2 | 1.2 | — |
gitleaksNot all datasets | 0.0 | 4.0 | 45.5 | — |
betterleaksNot all datasets | 0.0 | 0.2 | 44.8 | — |
gliner-multi-v21-ru+cpu-int8Not all datasets | — | 0.0 | — | — |
gliner-multi-v21+cpu-int8Not all datasets | 0.0 | 0.0 | 0.0 | — |
gliner-nvidia-ru+cpu-int8Not all datasets | — | 0.0 | — | — |
gliner-nvidia+cpu-int8Not all datasets | 0.0 | 0.0 | 0.0 | — |
gliner-urchade-ru+cpu-int8Not all datasets | — | 0.0 | — | — |
gliner-urchade+cpu-int8Not all datasets | 0.0 | 0.0 | 0.0 | — |
Personal data vs secrets
Compare how well each detector hides personal data, credentials, and other secrets.
DetectorPersonal dataSecrets
Use one detector or combine several?
See what combining detectors adds and how much extra text it masks.
One detector vs several combined
See whether combining detectors hides more sensitive data and how much extra text it masks.
| Detector setup | Completely hidden ↑ 0%50%100% | Extra text masked ↓ | Not fully hidden ↓ |
|---|---|---|---|
| Single detectorPPLX | 57,052 | ||
| Single detectorGLiNER2 Fastino | 45,819 | ||
| Two detectorsFastino + PPLX | 18,344 | ||
| Three detectorsFastino + PPLX + mmBERT | 16,011 | ||
| CPU + GPUFastino + PPLX + BardsAI EU | 13,471 | ||
| Four detectorsFastino + PPLX + BardsAI EU + mmBERT | 12,728 |
Single detectorRuns on CUDACPU + GPU
Find the best detector for each data type
Focus on the sensitive information your application needs to protect.
Hardest datasets
Higher scores mean the dataset was harder for the tested detectors.
rubai-ru
92.7Extreme
alexen2
73.0Extreme
multiconer-ru
67.5Extreme
corrupt-secrets-issues
64.2Hard
factrueval
57.8Hard
tab-echr
56.3Hard
secrets-issues
55.6Hard
synth-env-configs
50.5Hard
leak-museum
47.5Hard
corrupt-redmadrobot
46.6Hard
synth-secrets-en
42.0Medium
dialogpii-multi
40.9Medium
synth-secrets-ru
40.5Medium
leaky-repo
40.0Medium
privy
36.2Medium
corrupt-hivetrace
35.6Medium
redmadrobot
34.8Medium
tonicai
34.6Medium
synth-ru-tickets
33.4Medium
synth-jira-comments
32.9Medium
redact-multi
31.8Medium
synth-wiki-tables
31.4Medium
dialogpii-en
31.0Medium
nym-ru
30.3Medium
scanpatch
29.8Medium
nemotron-pii
29.6Medium
gretel-multi
29.4Medium
jayguard
28.0Medium
nym-multi
28.0Medium
ameau01
27.5Medium
nerel
25.9Medium
hivetrace
25.5Medium
kiji-multi
23.3Easy
redact-ru
23.3Easy
nym-en
22.0Easy
alrosait
21.5Easy
creddata
17.9Easy
kiji-en
15.6Easy
secrets-rules
12.6Easy
russian-pii-66k
10.2Easy
arthur-passwords
7.1Easy
Best detector by data type
Choose a type of sensitive data to see which detectors hide it best.
Detector leaderboard
31 detector setups · sorted by completely hidden
| Compare | # | Family | |||||||
|---|---|---|---|---|---|---|---|---|---|
| 01 | gliner2-fastino | GLINER2 | 79.86% | 86.81% | 5.88% | 0.763 | 14.93 | 0.27 | |
| 02 | pplx | PPLX | 74.92% | 79.05% | 11.42% | 0.716 | 9.94 | 2.20 | |
| 03 | nuner-zero | GLINER | 71.63% | 78.78% | 3.94% | 0.760 | 20.12 | 0.38 | |
| 04 | bardsai-eu | ONNX | 66.62% | 81.37% | 7.72% | 0.703 | — | — | |
| 05 | gliner2-large | GLINER2 | 65.65% | 71.98% | 3.51% | 0.685 | 24.69 | 0.61 | |
| 06 | gliner-urchade | GLINER | 65.51% | 72.52% | 3.79% | 0.687 | 9.68 | 0.20 | |
| 07 | gliner25-fastino | GLINER2 | 62.77% | 69.13% | 7.33% | 0.696 | 12.47 | 0.15 | |
| 08 | apararti | OPF | 62.49% | 70.23% | 11.16% | 0.658 | 4.09 | 1.69 | |
| 09 | openai-base | OPF | 59.69% | 65.39% | 8.83% | 0.657 | 4.21 | 1.80 | |
| 10 | gliner-pii-edge | GLINER | 57.10% | 68.30% | 9.54% | 0.629 | 6.88 | 5.80 | |
| 11 | opf-kz-ru | OPF | 56.51% | 65.16% | 10.90% | 0.634 | 3.39 | 1.59 | |
| 12 | pii-shield-onnx | ONNX | 56.43% | 69.64% | 38.92% | 0.416 | — | — | |
| 13 | ru-pii-ner | RUPII | 53.11% | 57.30% | 2.73% | 0.599 | 1.42 | — | |
| 14 | gliner-stream-pii | GLINER | 51.67% | 63.82% | 3.55% | 0.636 | 8.20 | 0.21 | |
| 15 | gliner-multi-v21 | GLINER | 51.28% | 60.10% | 2.71% | 0.537 | 9.89 | 0.19 | |
| 16 | mmbert32k | HF | 50.17% | 78.53% | 10.33% | 0.613 | 2.32 | 0.18 | |
| 17 | traciora | OPF | 48.60% | 59.54% | 7.51% | 0.638 | 3.29 | 1.36 | |
| 18 | opf-ru-v2 | OPF | 47.51% | 57.15% | 6.70% | 0.629 | 3.47 | — | |
| 19 | gliner2-vladlinv | GLINER2 | 46.09% | 49.61% | 0.91% | 0.572 | 14.49 | 0.17 | |
| 20 | gliner-pii-base | GLINER | 44.64% | 50.53% | 1.86% | 0.550 | 5.24 | 0.16 | |
| 21 | ru-legal-ner | HF | 44.15% | 65.45% | 12.56% | 0.449 | 0.20 | 0.13 | |
| 22 | davlan-xlmr | HF | 44.00% | 51.33% | 0.71% | 0.501 | 2.26 | 0.14 | |
| 23 | stanza-ru | STANZA | 43.41% | 51.15% | 16.11% | 0.326 | 8.35 | — | |
| 24 | gravitee-small | HF | 42.98% | 50.49% | 5.36% | 0.496 | 0.66 | — | |
| 25 | davlan-mbert | HF | 41.98% | 48.88% | 0.94% | 0.481 | 2.51 | 0.15 | |
| 26 | ner-ru-yqelz | HF | 39.74% | 49.75% | 9.17% | 0.445 | 7.08 | 0.18 | |
| 27 | fef2-secret-ru | HF | 35.60% | 43.35% | 2.17% | 0.530 | 3.05 | 0.15 | |
| 28 | ner-ru-gherman | HF | 27.63% | 48.78% | 0.28% | 0.453 | 2.48 | 0.15 | |
| 29 | spacy-ru-lg | SPACY | 26.75% | 31.36% | 3.25% | 0.326 | 0.19 | — | |
| 30 | rules-ru | RULES | 24.48% | 25.37% | 6.96% | 0.506 | — | — | |
| 31 | natasha | NATASHA | 24.15% | 28.03% | 1.07% | 0.316 | 0.08 | — |
How to read these numbers. Completely hidden means the whole sensitive value was masked. Found at least partly can still leave text visible. Extra text masked is measured on unlabeled rows and is not a verified false-positive rate.
Benchmark size
What was tested
These counts describe the test data; they do not show which detector is best.
Detector setups tested10362 catalog records · variants included
Languages3Russian · English · Multilingual
Text rows tested41,6432,591 saved prediction runs overall
Labeled sensitive items227,466Values expanded to word boundaries