SiperBook a demo
Menu

Benchmark · October 2026

Siper finds 85.7% of the personal data in our Turkish test set. The nearest alternative finds 69.3%.

Measured 2 November 2026, "Turkish and English" option. Every system ran on the same 403 documents.

To read these numbers together, see how to read a personal data benchmark. It explains recall, precision and false alarm.

Special categories

On health data, Siper finds 81%. The nearest alternative finds 51%.

Siper has the highest recall in 5 of 7 special categories under KVKK Article 6. Siper is the only system above 0.65 in all 7 categories.

Recall of Siper and of the nearest alternative in each special category
CategorySiperNearest alternative
Health81%Higher51%GPT-5-mini
Religion86%88%GPT-5-mini, higher
Ethnic origin91%96%GPT-5-mini, higher
Political opinion88%Higher81%GPT-5-mini
Union membership94%Higher89%GPT-5-mini
Criminal conviction77%Higher53%GPT-5-mini
Sex life90%Higher71%GPT-5-mini

By class

Siper's recall in each class

Measured 2 November 2026, "Turkish and English" option.

Siper recall for person name, address and seven special-category classes
ClassRecall
Person name90.4%
Address77.8%
Health80.7%
Religion86.0%
Ethnic origin91.2%
Political opinion88.1%
Union membership94.0%
Criminal conviction77.4%
Sex life89.6%

By category

Only 4 of 10 systems have a label for all 7 special categories.

Each cell shows recall in one category. Darker cells mean higher recall. A cell marked n/a is out of scope: the system has no label for that category. A cell with 0% is a measured zero, because the system has the label and found nothing correct on this set.

On a phone, scroll the table sideways.

Showing 10 of 10 systems in both tables. Showing 4 of 10 systems in both tables.

Recall per system in each of the 7 special categories, highest health recall first
SystemHealthReligionEthnic originPolitical opinionUnion membershipCriminal convictionSex life
Gurubase Siper81%86%91%88%94%77%90%
GPT-5-mini51%88%96%81%89%53%71%
Limina (Private AI)42%56%91%37%n/a0%4%
Qwen2.5-32B37%30%79%47%85%17%56%
Gemini 2.5 Flash24%78%68%39%77%15%52%
Google Cloud DLP0%30%60%0%0%6%n/a
Tonic Textual0%n/a9%n/an/an/an/a
Azure AI Languagen/an/an/an/an/an/an/a
OpenAI Privacy Filtern/an/an/an/an/an/an/a
Microsoft Presidion/an/an/an/an/an/an/a

Systems

Siper has the highest recall and F1.

Among systems that can run on site, the nearest alternative is Qwen2.5-32B, with recall of 48.3%.

The tables list 9 of the 15 alternatives we evaluated. Ranges are 95% confidence intervals from 1,000 document resamples. Range values and bars use a scale from 0 to 1.

Sorted by recall, highest first. On a phone, scroll the table sideways.

Showing 10 of 10 systems in both tables. Showing 4 of 10 systems in both tables.

Recall, precision, F1, false alarm rate and latency for each system
SystemCan run on siteRecallPrecisionF1False alarmMedian latency
Gurubase SiperYes85.7%0.839 to 0.87756.1%0.6780.652 to 0.70014.3%176 msOne 8-vCPU server, no GPU · October 2026
GPT-5-miniNo69.3%0.653 to 0.73062.8%0.6590.632 to 0.68412.1%18,115 msp95 19,621 msCloud API · July 2026
Gemini 2.5 FlashNo52.2%0.470 to 0.57161.5%0.5650.526 to 0.6001.9%1,790 msp95 2,072 msCloud API · July 2026
Limina (Private AI)No51.5%0.477 to 0.55265.2%0.5750.543 to 0.60548.6%561 msp95 629 msCloud API · July 2026
Qwen2.5-32BYes48.3%0.446 to 0.52158.5%0.5290.496 to 0.561Not measured19,242 msp95 19,367 msLocal GPU · July 2026
Azure AI LanguageNo26.7%0.220 to 0.31191.8%0.4140.355 to 0.4663.4%221 msp95 264 msCloud API · July 2026
Google Cloud DLPNo24.5%0.202 to 0.28842.6%0.3110.265 to 0.35440.2%220 msp95 289 msCloud API · July 2026
Tonic TextualNo22.2%0.180 to 0.26524.1%0.2310.193 to 0.26442.7%491 msp95 568 msCloud API · July 2026
Microsoft PresidioYes19.2%0.156 to 0.2299.0%0.1220.098 to 0.14698.1%7 msp95 8 msLocal CPU · July 2026
OpenAI Privacy FilterYes8.9%0.059 to 0.12048.9%0.1500.103 to 0.19818.4%4,600 msOne 8-vCPU server, no GPU · 2026
Recall
The share of real personal data that a system finds.
Precision
The share of detections that are correct. Siper's precision is 56.1%.
F1
One score that combines recall and precision.
False alarm
The share of 321 real documents with no personal data that received at least one unnecessary mask. Read it with recall, because a system that detects little also makes few false alarms.
Median latency
The middle response time. On-prem systems: median per document on one 8-vCPU server without GPU, unless the row states another setup. Cloud services: API response time including network. Systems ran on different infrastructure, so this is not a direct hardware comparison. Dates also differ, as the table shows. Use it to compare orders of magnitude.

Method

Every system ran on the same documents and the same scoring software.

  • The test set has 403 Turkish documents with human-verified labels. We evaluated the alternatives in October 2026. Siper's own numbers are from 2 November 2026.
  • A referee review corrected the test set. We then rescored all systems on the corrected set.
  • The scoring software runs independently of the compared products. It measures at entity level.
  • Commercial products ran through their official APIs, with real access keys, on the evaluation date.
  • We mapped each system's labels to one common taxonomy. A category that a system never claims counts as out of scope, not as zero.
  • Each score carries a 95% confidence interval from 1,000 document resamples.
  • False alarms come from a separate group of 321 real documents with no personal data.
  • We record the date and the system version of each evaluation.

We ran 135 paired comparisons: 15 alternatives across 9 classes (person name, address and seven special-category classes). We used document-level paired resampling, and we claim no lead when the interval of the difference includes zero. 113 comparisons showed a significant difference in Siper's favour, and 20 showed no significant difference. 2 favoured the alternative.

Latency: On-prem systems: median per document on one 8-vCPU server without GPU, unless the row states another setup. Cloud services: API response time including network. Systems ran on different infrastructure, so this is not a direct hardware comparison.

Not in the tables: Global cloud provider C. Its official documentation states that it does not support Turkish personal data detection.

A detailed report with mapping rules, system configurations and rerun notes is available under a non-disclosure agreement. The JSON file carries the status and the source of the data.

Ask for the reportRead these results as JSON

Limits

Limits of this test

Read the results with these limits in mind.

  • The test set is ours, and this is not an independent audit.
  • The test set is Turkish only. Siper's accuracy figures use its "Turkish and English" option. Siper offers three language options: read how they differ.
  • Systems ran on different infrastructure, so use the table to compare orders of magnitude.
  • Prompt injection detection is not part of this test.

Next step

Test Siper on your own text.