# euRedact Cloud on the Benchmark, per Type

> euRedact Cloud and four other systems on the published 900-document benchmark in five languages: what each finds and leaves readable per type of personal data, per set and per language.

Markdown version of https://euredact.dev/benchmarks/cloud, generated from that page at build on 2026-11-19. Every page of the site is listed with its Markdown version at https://euredact.dev/llms.txt; the full SDK reference is https://euredact.dev/llms-full.txt.

*Benchmarks*

# euRedact Cloud Benchmarked

euRedact Cloud and four other systems on the published benchmark: 900 synthetic IT, HR and medical documents in five languages, 27,831 annotated items of personal data. What each system finds, and what it leaves readable.

[The benchmark on GitHub →](https://github.com/euredact/benchmark-2026-11-14)[The launch post →](/blog/euredact-cloud-public-beta)[The rules engine’s figures →](/benchmarks)

## All 900 documents

F2 counts recall twice, because a missed item leaks. “Found” is recall with overlap matching: an item counts when any part of it was redacted. “Left readable” is the share of items with less than 75% of their letters and digits hidden. The Article 9 and 10 column is the share of those items masked by the same strict rule.

*System*

*Measured*

*F2*

*Found*

*Precision*

*Left readable*

*Art. 9/10 masked*

| --- | --- | --- | --- | --- | --- | --- |
| euRedact | 2026-11-14 | 0.921 | 94.2% | 84.8% | 7.3% | 78.5% |
| GLiNER2-PII | 2026-11-02 | 0.706 | 68.7% | 70.3% | 33.0% | 1.9% |
| MS Presidio | 2026-11-02 | 0.656 | 63.2% | 77.2% | 44.6% | 14.1% |
| bards.ai v3 | 2026-11-02 | 0.632 | 58.7% | 71.0% | 43.9% | 0.7% |
| Mistral Small 4 | 2026-11-02 | 0.561 | 53.0% | 73.4% | 48.5% | 5.5% |

F2 is each system’s figure as published in the benchmark’s held-out review; the other columns pool the three sets’ counts. Pooled the same way, F2 is 0.690 for GLiNER2-PII and 0.608 for bards.ai v3. euRedact’s release frozen before these sets were opened (1 November, measured 2 November with the other four) scored F2 0.919, found 82% of the Article 9 and 10 items and left 7.8% readable; the table shows the release of 10 November.

## Per type of personal data

The benchmark’s own 35 types, grouped as it groups them, with the Article 9 and 10 data first. “Items” is how many of that type the 900 documents hold; the other four systems are shown by what they found. GLiNER2-PII was not asked for the Article 9 and 10 types.

*Type*

*Items*

*euRedact*

*Found by the four others*

| --- | --- | --- | --- |

*Found*

*Left readable*

*GLiNER2-PII*

*MS Presidio*

*bards.ai v3*

*Mistral Small 4*

| --- | --- | --- | --- | --- | --- |

*Article 9 and 10: special categories and offences*

| HealthHEALTH | 5,188 | 86.9% | 17.9% | 1.6% | 37.1% | 0.9% | 10.3% |
| Criminal offenceCRIMINAL\_OFFENCE | 362 | 64.6% | 41.7% | 2.8% | 39.0% | 0.8% | 5.8% |
| Genetic dataGENETIC | 236 | 89.4% | 17.8% | 6.8% | 57.2% | 3.4% | 20.8% |
| Religion or beliefRELIGION\_BELIEF | 211 | 80.6% | 28.0% | 10.4% | 63.5% | 8.5% | 15.6% |
| Biometric referenceBIOMETRIC | 207 | 61.8% | 47.8% | 26.6% | 64.7% | 9.2% | 20.3% |
| Trade-union membershipTRADE\_UNION | 177 | 86.4% | 22.0% | 18.1% | 77.4% | 3.4% | 1.7% |
| Political opinionPOLITICAL\_OPINION | 154 | 78.6% | 34.4% | 22.1% | 40.9% | 9.7% | 8.4% |
| Sex life or orientationSEX\_LIFE\_ORIENTATION | 143 | 48.9% | 53.8% | 2.8% | 46.2% | 0.7% | 10.5% |
| Ethnic originETHNIC\_ORIGIN | 134 | 91.8% | 11.9% | 27.6% | 73.9% | 20.2% | 25.4% |

*Names*

| Person namePERSON | 7,034 | 99.4% | 0.7% | 94.9% | 93.6% | 97.2% | 66.5% |
| Username or handleUSERNAME | 883 | 95.3% | 4.8% | 88.8% | 62.8% | 62.5% | 64.9% |

*Occupation*

| Job titleJOB\_TITLE | 3,244 | 96.9% | 3.6% | 84.8% | 10.2% | 23.5% | 25.5% |

*Places*

| CityCITY | 1,565 | 96.7% | 3.3% | 96.4% | 99.0% | 83.3% | 68.2% |
| Street addressSTREET\_ADDRESS | 655 | 100.0% | 0.1% | 98.9% | 99.9% | 93.1% | 94.7% |
| PostcodePOSTCODE | 645 | 100.0% | 0.3% | 99.4% | 98.1% | 92.6% | 95.2% |

*Contact details*

| E-mail addressEMAIL | 903 | 100.0% | 0.0% | 99.6% | 99.0% | 100.0% | 88.6% |
| Phone numberPHONE | 407 | 100.0% | 0.3% | 100.0% | 94.3% | 99.0% | 96.3% |

*Government identifiers*

| Tax identifierTAX\_ID | 246 | 100.0% | 1.2% | 83.3% | 78.5% | 97.6% | 88.2% |
| Vehicle plateVEHICLE\_PLATE | 144 | 94.4% | 5.6% | 36.1% | 73.6% | 94.4% | 93.8% |
| National identifierNATIONAL\_ID | 135 | 100.0% | 0.0% | 84.4% | 88.1% | 96.3% | 95.6% |
| Passport numberPASSPORT | 127 | 100.0% | 0.0% | 98.4% | 89.0% | 96.1% | 97.6% |
| Driving licenceDRIVERS\_LICENSE | 121 | 100.0% | 0.0% | 91.7% | 86.0% | 92.6% | 95.9% |

*Financial*

| Financial amountFINANCIAL\_AMOUNT | 1,901 | 95.8% | 4.4% | 90.8% | 0.5% | 54.1% | 55.8% |
| IBANIBAN | 212 | 100.0% | 0.0% | 91.5% | 99.1% | 96.2% | 93.4% |
| Payment cardPAYMENT\_CARD | 151 | 100.0% | 0.0% | 27.2% | 99.3% | 24.5% | 80.1% |

*Network identifiers*

| IP addressIP\_ADDRESS | 523 | 100.0% | 0.0% | 99.2% | 99.6% | 96.8% | 76.9% |
| MAC addressMAC\_ADDRESS | 212 | 100.0% | 0.0% | 25.0% | 100.0% | 90.6% | 82.1% |
| Personal URLPERSONAL\_URL | 170 | 100.0% | 0.6% | 0.0% | 98.2% | 9.4% | 90.6% |

*Other identifiers*

| Employee identifierEMPLOYEE\_ID | 531 | 96.2% | 4.3% | 86.1% | 69.7% | 83.0% | 88.5% |
| Device serialDEVICE\_SERIAL | 236 | 25.4% | 74.6% | 42.4% | 60.2% | 67.8% | 83.5% |
| Customer identifierCUSTOMER\_ID | 187 | 98.4% | 1.6% | 83.4% | 57.8% | 86.1% | 88.2% |
| Professional identifierPROFESSIONAL\_ID | 166 | 100.0% | 0.0% | 65.1% | 81.3% | 93.4% | 95.8% |
| Medical record numberMEDICAL\_RECORD\_NUMBER | 143 | 100.0% | 0.0% | 55.3% | 76.2% | 97.9% | 99.3% |

*Credentials*

| Secret or API keySECRET | 155 | 99.4% | 7.1% | 89.7% | 71.6% | 72.3% | 91.6% |

*Dates*

| Date of birthDATE\_OF\_BIRTH | 323 | 100.0% | 0.6% | 100.0% | 93.2% | 99.7% | 99.1% |

## Per set and per language

Three sets of 300 documents, and sixty documents per language in each set. F2, with the share left readable under it.

*System*

*Per set*

*Per language*

| --- | --- | --- |

*IT*

*HR*

*Medical*

*German*

*English*

*French*

*Dutch*

*Polish*

| --- | --- | --- | --- | --- | --- | --- | --- |
| euRedact | 0.9315.8% readable | 0.9455.6% readable | 0.9009.5% readable | 0.9295.9% readable | 0.9197.2% readable | 0.9247.6% readable | 0.9118.7% readable | 0.9247.2% readable |
| GLiNER2-PII | 0.80916.5% readable | 0.78920.7% readable | 0.52752.2% readable | 0.68732.7% readable | 0.70230.4% readable | 0.67835.0% readable | 0.69732.8% readable | 0.68633.9% readable |
| MS Presidio | 0.70733.0% readable | 0.68938.4% readable | 0.59856.7% readable | 0.67040.7% readable | 0.65547.1% readable | 0.65045.7% readable | 0.66442.7% readable | 0.64047.2% readable |
| bards.ai v3 | 0.74127.8% readable | 0.69234.9% readable | 0.44960.8% readable | 0.60743.5% readable | 0.63340.3% readable | 0.60944.6% readable | 0.61043.3% readable | 0.58147.8% readable |
| Mistral Small 4 | 0.64537.8% readable | 0.61241.7% readable | 0.46560.2% readable | 0.56348.0% readable | 0.56448.1% readable | 0.55748.9% readable | 0.55249.5% readable | 0.56948.0% readable |

## How to read these figures

**When and how.** euRedact was measured on 14 November 2026 through the production service with the public SDK 1.0.0 (model release of 10 November). The other four were measured on 2 November 2026, with every configuration fixed before the sets were opened. One run per system.

**The data.** Synthetic documents, no customer data, written and annotated by language models under one guideline family; agreement before adjudication was 0.85–0.97. None of these documents is in euRedact’s training data. The answer key’s guideline is not euRedact’s own set of definitions: where the two disagree, euRedact loses points.

**The other systems.** MS Presidio with EU recognizers and term lists, configured in six rounds against 65 development documents from the benchmark’s earlier sets, none of them among the 900. GLiNER2-PII with its model card’s labels plus job title and money amount. Mistral Small 4 with a prompt that lists its 34 types, each with a definition, and asks for structured output. bards.ai v3 decoded as its model card does.

**Which systems.** Compared systems are limited to those that can be self-hosted or are operated under EU jurisdiction only. Services subject to the US CLOUD Act are out of scope.

**Precision.** Part of what euRedact redacts beyond the annotations is masked on purpose, such as departments and tests in a medical context; that is one of the reasons its precision is 85% rather than higher.

**The all-sets F2** is the figure the benchmark publishes for each system. Pooling the three sets’ counts gives 0.690 for GLiNER2-PII and 0.608 for bards.ai v3; the per-set and per-type figures are exact.

**Rerun it.** A 60-document sample with its gold standard, the harness, the configurations and our scores are [on GitHub](https://github.com/euredact/benchmark-2026-11-14) (source-available, for internal use). All 900 documents and their annotations are pinned by hash, and euRedact’s full-set predictions are committed by hash, so they can’t be changed after publication. The figures on this page were exported from the benchmark repository at commit 92863c1 on 2026-11-16.

## Run It on Your Documents

euRedact was measured through the same service a free account uses. The harness on GitHub takes an API key and runs the 60-document sample on the free plan; your own documents go in the same way, and no text or result is kept. Or see a redaction first in the live demo, on a sample or text you made up, no account needed.

[Try the live demo](/demo)[The sample and harness on GitHub →](https://github.com/euredact/benchmark-2026-11-14)
