Skip to content
HireLayer

HireLayer vs HrFlow.ai, on 30 resumes neither had seen

Tested in October 2026. The 30 resumes print 1,886 values: names, job titles, dates, schools, skills, languages. We checked how many of them each parser returned exactly as the resume prints them.

  1. HireLayer

    CV Extract

    95.2%

  2. HrFlow.ai Gaia

    Most accurate model

    84.5%

  3. HrFlow.ai Hawk

    Default model

    73.3%

Values read correctly, out of 1,886
Migration from HrFlow.aiDownload the data

Score by category

Share of the values printed on the 30 resumes that each parser returned as printed. Darker means higher. HireLayer names skills in English when the resume is not in French, so a blind reviewer checked those matches without knowing wich parser wrote them.

Share of printed values each parser returned exactly, by category
CategoryHireLayerGaiaHawk
Contact details187 values · Name, email, phone, city, postal code, LinkedIn97.9%88.2%87.7%
Work experience834 values · Title, employer, dates, location, description95.1%89.0%81.8%
Education254 values · Degree, school, dates90.6%95.7%77.2%
Listed skills288 values · Skills printed in the skills section96.2%82.3%54.9%
Languages171 values · Language and level95.9%48.0%50.3%
Certifications69 values · Licences and certificates97.1%69.6%36.2%
Interests61 values · Hobbies and interests100.0%93.4%98.4%
Birth date, driving licence22 values · When printed95.5%86.4%54.5%
All dates403 values · Start and end of every position and diploma97.3%84.9%83.9%

Where the difference comes from

HireLayer reads the resumes more accurately. The two APIs also return different fields.

Reading quality

On the 1,764 values both APIs return, so the only difference is how well each one reads the resume.

HireLayer
95.3%
Gaia
88.9%
Hawk
77.4%

  • Job datesMonth and year of each position

    • HireLayer235/244
    • Gaia203/244
    • Hawk212/244

  • Listed skillsSkills printed in the skills section

    • HireLayer277/288
    • Gaia237/288
    • Hawk158/288

  • CertificationsLicences and certificates

    • HireLayer67/69
    • Gaia48/69
    • Hawk25/69

  • Postal codeAs printed

    • HireLayer21/21
    • Gaia15/21
    • Hawk15/21

  • LinkedIn URLFull profile address

    • HireLayer13/13
    • Gaia8/13
    • Hawk8/13

  • CityCandidate's city

    • HireLayer28/30
    • Gaia23/30
    • Hawk23/30

  • Job titleAs printed

    • HireLayer131/141
    • Gaia132/141
    • Hawk107/141

  • EmployerCompany name

    • HireLayer123/137
    • Gaia123/137
    • Hawk87/137

  • Degree titleAs printed

    • HireLayer52/67
    • Gaia65/67
    • Hawk44/67

  • SchoolSchool or university

    • HireLayer59/66
    • Gaia63/66
    • Hawk43/66

Gaia was ahead on degree titles and schools, and level with HireLayer on job titles and employers. HireLayer led on dates, skills, certifications and contact details.

Different fields

Each API decides what it returns. The scores at the top include the two fields below, the reading score on the left leaves them out.

HireLayer returns, HrFlow.ai does not

  • Language level

    HireLayer returns the level of each language. HrFlow.ai returns the language without its level.

    • HireLayer

      77/84

    • Gaia

      0/84

    • Hawk

      0/84

  • Current job

    HireLayer marks the job the candidate still holds as current. HrFlow.ai often gives it an end date.

    • HireLayer

      38/38

    • Gaia

      24/38

    • Hawk

      17/38

HrFlow.ai returns, HireLayer does not

  • A short summary of the profile
  • A list of tasks for each job
  • Skills attached to each job
  • A gender field, filled by Gaia on a few resumes

Evidence from French and English resumes

Values as printed, and as each parser returned them.

Reading errors

cv18.pdfEnglish · table, dates on the right

On the resume

Assistant General Manager – Warehouse Operations
Apr'21 – Till date
  • HireLayer2021-04 → current
  • Gaia2021-01 → 2021-04
  • Hawk2015-01 → 2024-12
cv19.pdfEnglish · left sidebar

On the resume

linkedin.com/in/adaeze-okonkwo-rm-3b09
  • HireLayerfull URL
  • Gaialinkedin.com/in/
  • Hawklinkedin.com/in/
cv15.pdfEnglish · scanned PDF

On the resume

Skills: Dementia care · Palliative care · Infection prevention and control · Care documentation
  • HireLayerall 4
  • Gaianone
  • Hawknone

Different fields

cv06.pdfFrench · table, dates on the right

On the resume

Puéricultrice (4/5 temps) · Crèche Les Coccinelles
sept. 2017 – à ce jour
  • HireLayer2017-09 → current
  • Gaia2017-01 → 2017-09
  • Hawk2017-01 → 2017-09
cv01.pdfFrench · dates under the title

On the resume

Français – langue maternelle · Arabe – courant (oral)
  • HireLayerlevel for each
  • Gaiano level
  • Hawkno level
cv13.docxEnglish · Word, header contact

On the resume

English – Native · Urdu – Fluent (spoken) · Punjabi – Conversational
  • HireLayerlevel for each
  • Gaiano level
  • Hawkno level

All 30 resumes

One line per resume, three marks per line. The further right, the more of its printed values the parser returned. Filter by language or file type, and click a resume number to download it.

HireLayer average
95.3%
Gaia average
84.0%
Hawk average
72.9%
HireLayerHrFlow.ai GaiaHrFlow.ai HawkShare of each resume's printed values returned exactly
  1. cv01Night-shift care assistant, nursing homeFrench · France · PDF · Dates under the title

    HireLayer 93.9%. HrFlow.ai Gaia 89.8%. HrFlow.ai Hawk 53.1%.

  2. cv02Construction works managerFrench · France · PDF · Two-block design

    HireLayer 98.8%. HrFlow.ai Gaia 86.2%. HrFlow.ai Hawk 80.0%.

  3. cv03Assistant store manager, fashion retailFrench · France · DOCX · Word, contact in page header

    HireLayer 100.0%. HrFlow.ai Gaia 88.9%. HrFlow.ai Hawk 82.2%.

  4. cv04Public procurement officer, local governmentFrench · France · DOCX · Word, one column

    HireLayer 95.6%. HrFlow.ai Gaia 82.4%. HrFlow.ai Hawk 72.1%.

  5. cv05Psychology student, part-time jobsFrench · France · PDF · Two text columns

    HireLayer 90.7%. HrFlow.ai Gaia 85.2%. HrFlow.ai Hawk 70.4%.

  6. cv06Nursery nurse, LiègeFrench · Belgium · PDF · Table, dates on the right

    HireLayer 100.0%. HrFlow.ai Gaia 87.3%. HrFlow.ai Hawk 63.6%.

  7. cv07Watchmaker, cross-border workerFrench · Switzerland · DOCX · Word, contact in page header

    HireLayer 94.6%. HrFlow.ai Gaia 89.3%. HrFlow.ai Hawk 67.9%.

  8. cv08Personal finance advisor, QuebecFrench · Canada · PDF · Right sidebar

    HireLayer 90.9%. HrFlow.ai Gaia 81.8%. HrFlow.ai Hawk 69.1%.

  9. cv09Chief financial officer, food groupFrench · France · PDF · One column

    HireLayer 90.1%. HrFlow.ai Gaia 93.4%. HrFlow.ai Hawk 73.6%.

  10. cv10Fire safety officer, temp contractsFrench · France · Scanned JPG · One column

    HireLayer 94.7%. HrFlow.ai Gaia 91.2%. HrFlow.ai Hawk 73.7%.

  11. cv11Restaurant general managerEnglish · United States · PDF · US one-page resume

    HireLayer 98.1%. HrFlow.ai Gaia 94.2%. HrFlow.ai Hawk 82.7%.

  12. cv12HVAC service technicianEnglish · United States · Phone photo · Dates under the title

    HireLayer 86.4%. HrFlow.ai Gaia 81.8%. HrFlow.ai Hawk 70.5%.

  13. cv13Contact centre team leaderEnglish · United Kingdom · DOCX · Word, contact in page header

    HireLayer 98.3%. HrFlow.ai Gaia 91.5%. HrFlow.ai Hawk 78.0%.

  14. cv14Frontend engineer, DublinEnglish · Ireland · PDF · Two-block design

    HireLayer 98.6%. HrFlow.ai Gaia 95.8%. HrFlow.ai Hawk 94.4%.

  15. cv15Personal support workerEnglish · Canada · Scanned PDF · Table, dates on the right

    HireLayer 91.5%. HrFlow.ai Gaia 83.1%. HrFlow.ai Hawk 69.5%.

  16. cv16Student, casual hospitality jobsEnglish · Australia · PDF · Dates under the title

    HireLayer 100.0%. HrFlow.ai Gaia 77.4%. HrFlow.ai Hawk 60.4%.

  17. cv17Dairy farm managerEnglish · New Zealand · DOCX · Word, one column

    HireLayer 88.6%. HrFlow.ai Gaia 89.8%. HrFlow.ai Hawk 62.5%.

  18. cv18Warehouse operations manager (3PL)English · India · PDF · Table, dates on the right

    HireLayer 88.3%. HrFlow.ai Gaia 87.0%. HrFlow.ai Hawk 83.1%.

  19. cv19Corporate banking relationship managerEnglish · Nigeria · PDF · Left sidebar

    HireLayer 98.6%. HrFlow.ai Gaia 93.0%. HrFlow.ai Hawk 76.1%.

  20. cv20Cabin crew moving into L&DEnglish · Singapore · PDF · Functional resume

    HireLayer 98.4%. HrFlow.ai Gaia 90.5%. HrFlow.ai Hawk 69.8%.

  21. cv21MIG/MAG/TIG welderPolish · Poland · DOCX · Word, two columns

    HireLayer 98.3%. HrFlow.ai Gaia 70.7%. HrFlow.ai Hawk 70.7%.

  22. cv22Customer service specialist with FrenchPolish · Poland · DOCX · Word, two columns

    HireLayer 96.1%. HrFlow.ai Gaia 74.5%. HrFlow.ai Hawk 64.7%.

  23. cv23Preschool teacherSwedish · Sweden · PDF · Header banner

    HireLayer 91.9%. HrFlow.ai Gaia 79.0%. HrFlow.ai Hawk 75.8%.

  24. cv24Chemical engineering lecturerPortuguese · Portugal · PDF · Academic CV, 4 pages

    HireLayer 95.2%. HrFlow.ai Gaia 3.2%. HrFlow.ai Hawk 69.4%.

  25. cv25Embedded automotive software engineerRomanian · Romania · PDF · Europass

    HireLayer 100.0%. HrFlow.ai Gaia 95.7%. HrFlow.ai Hawk 85.7%.

  26. cv26Junior graphic designerTurkish · Turkey · PDF · Infographic

    HireLayer 96.3%. HrFlow.ai Gaia 75.9%. HrFlow.ai Hawk 74.1%.

  27. cv27Plant director, automotive supplierCzech · Czechia · DOCX · Word tables

    HireLayer 96.2%. HrFlow.ai Gaia 91.2%. HrFlow.ai Hawk 85.0%.

  28. cv28Front office manager, hotelGerman · Austria · PDF · Dates on the left

    HireLayer 98.4%. HrFlow.ai Gaia 90.6%. HrFlow.ai Hawk 79.7%.

  29. cv29Physiotherapist, hospital and private practiceDutch · Belgium · PDF · Dates under the title

    HireLayer 98.8%. HrFlow.ai Gaia 85.7%. HrFlow.ai Hawk 63.1%.

  30. cv30Insurance sales advisorSpanish · Colombia · PDF · Two text columns

    HireLayer 90.7%. HrFlow.ai Gaia 94.4%. HrFlow.ai Hawk 66.7%.

Migration from HrFlow.ai, without rewriting your integration

HireLayer answers in HrFlow.ai's format. The JSON your application recieves keeps the same shape, so your mapping, your database and your screens stay as they are. There are two ways to do it.

You use HrFlow.ai's SDK or API

Change the base URL and the key. For now this covers synchronous parsing calls only (sync_parsing=1).

parse.py

from hrflow import Hrflow client = Hrflow(    api_url="https://api.hrflow.ai/v1/",    api_url="https://hirelayer.co/api/hrflow/v1/",    api_secret=HRFLOW_API_KEY,    api_secret=HIRELAYER_API_KEY,    api_user="you@company.com",) response = client.profile.parsing.add_file(    source_key="my-source", profile_file=pdf, sync_parsing=1,)profile = response["data"]["profile"]

You call HireLayer directly

Add one parameter to CV Extract to get HrFlow.ai's JSON.

terminal

curl -X POST \  "https://hirelayer.co/api/v3/parser" \  "https://hirelayer.co/api/v3/parser?output_format=hrflow" \  -H "X-API-Key: $HIRELAYER_API_KEY" \  -F "file=@resume.pdf" # 201 Created{ "code": 201, "message": "Profile parsed successfully.",  "data": { "parsing": {…}, "profile": {…} } }
Same objects
Profile and Parsing, with info, experiences, educations, skills, languages and certifications.
Real time
Every parse answers 201 with the profile inside. No queue to poll, no webhook to wait for.
Your identifiers
reference, tags, metadatas and labels come back in the profile.
Public prices
One credit per resume, and 50 free credits a month to compare on your own files.
Get your free API keyPlan your migration with us

Need fields HrFlow.ai's format has no place for, such as contract types or ROME occupation codes? Leave out the parameter and you get the full CV Extract response.

Protocol

HireLayer is our product, so we publish everything: the resumes, the expected values, the raw answers and the scoring code. All three parsers received the same files at the same time, from the same machine.

  1. 30 resumes written

    Fictional people, so every right answer is known. Written by people who saw no parser output.

  2. Rules frozen

    Expected values and scoring code fixed, their SHA-256 recorded, before any parser received a file.

  3. Same files, same moment

    Each file sent to HireLayer and to both HrFlow.ai Sources in parallel, twice.

  4. One logged fix

    Languages returned under their printed name now match. Its effect on each parser is published.

  5. Raw answers kept

    Every JSON answer saved, account identifiers removed, nothing else changed.

  6. Blind review

    Skills that string matching could not settle were judged with the parsers shown as A, B and C.

How each parser was called
HireLayer CV ExtractPOST /api/v3/parserSynchronous, do_not_store_data=true
HrFlow.ai GaiaPOST /v1/profile/parsing/fileSource with Gaia, presented as HrFlow.ai’s most accurate model
HrFlow.ai HawkPOST /v1/profile/parsing/fileSource with Hawk, the default model

Files and data

The archive lets you check the results or run the benchmark again.

  • hirelayer-vs-hrflow-benchmark/
  • corpus/cv/cv01…cv30.json# what each resume prints
  • results/raw/# 3 parsers × 2 runs, as returned
  • results/scores/# one row per scored value
  • results/judgements/# blind skill review
  • results/SCORING_FREEZE.md# hashes recorded before the calls
  • scripts/# render, call, score, review
  • METHODOLOGY.md# every rule and threshold
Data archive · ZIP, 1.8 MBGround truth, raw answers, scores, review, scripts

The 30 resume files download one by one from the “All 30 resumes” section: click a resume number. Every person is fictional. Emails use example domains and phone numbers come from ranges reserved for fiction or unasigned blocks.

Questions

Which parser was more accurate, HireLayer or HrFlow.ai?
On 30 resumes neither had seen, HireLayer returned 95.2% of the 1,886 printed values exactly, HrFlow.ai Gaia 84.5% and HrFlow.ai Hawk 73.3%. Leaving out language levels and current jobs, which HrFlow.ai does not return in the same way, the scores are 95.3%, 88.9% and 77.4%.
Which HrFlow.ai models were tested?
Gaia, which HrFlow.ai presents as its most accurate parsing model, and Hawk, the model selected by default when you create a Source. Each ran on its own Source with the same files as HireLayer, at the same time.
What makes these resumes “unseen”?
They were written after the 15 October 2026 HireLayer release, by writers who saw no parser output, and every person on them is fictional, so each right answer is known in advance. The expected values and the scoring code were frozen, with their SHA-256 recorded, before any parser received a file.
Does HrFlow.ai return language levels?
Not in this test. HrFlow.ai returned the languages without their level. HireLayer returns the level of each language, whether the resume writes it as a word, such as “courant” or “Fluent”, or as a grade.
How do I move from HrFlow.ai to HireLayer?
If you use HrFlow.ai’s SDK or API, change the API URL to https://hirelayer.co/api/hrflow/v1/ and use your HireLayer key: the parsing call and the JSON it returns keep HrFlow.ai’s format. For now this covers synchronous parsing calls only (sync_parsing=1). If you call HireLayer directly, add output_format=hrflow to the CV Extract request.
Can I check the results myself?
Yes. Download the 30 resumes and the data archive from this page: it holds what each resume prints, the raw JSON of the three parsers on both runs, the score of every value and the scripts that produced them.

Try it on your own resumes

Send a sample of the CVs you actually receive to HireLayer in HrFlow.ai's format and compare the profiles with the ones you have today.

Book a demoFeature comparison