Compare detectors
Compare on the same datasets, with the same denominator.
Dataset slice
Choose up to four detectors
Comparison uses only datasets available to every selected detector. The current selection is shareable from the URL.
41 shared datasets · 227,466 normalized gold annotations · Known training-source overlaps excluded
gliner2-fastino
79.86%fully hidden
Detected86.81%
Extra masking5.88%
Character F10.763
CPU s / 10k chars14.93
Untouched13.19%
Gold annotations227,466
pplx
74.92%fully hidden
Detected79.05%
Extra masking11.42%
Character F10.716
CPU s / 10k chars9.94
Untouched20.95%
Gold annotations227,466
Coverage vs extra masking
Same shared dataset intersection
Characters masked in unannotated rows →
Coverage vs CPU speed
Speed uses the fixed hardware workload; quality uses the shared dataset intersection
Measured CPU throughput · characters / second →
Full hiding by data type
Same dataset intersection for every column. Each cell retains its own annotation count.
| Data type | Gold annotations | gliner2-fastino | pplx |
|---|---|---|---|
| Passwords, keys & tokens | 6,870 | ||
| Logins & usernames | 11,382 | ||
| Bank accounts & cards | 5,253 | ||
| Documents & identifiers | 15,034 | ||
| People's names | 65,736 | ||
| Phone numbers & email | 21,981 | ||
| Addresses & locations | 39,602 | ||
| Dates & times | 5,502 | ||
| Organizations | 19,274 | ||
| Network identifiers | 25,631 | ||
| Customer & employee IDs | 7,994 | ||
| Other sensitive attributes | 3,207 |