Continue → Overview
← All research

Gender Prediction API Benchmark 2026

On a fixed fixture of 4,610 labelled names across 25 countries, NameGender answered 87.87% of rows, was correct on 96.03% of answered rows, and returned a correct answer for 84.38% of all rows. The country and script tables below are generated from the published report.

Measured August 28, 2026 · fixed fixture · fuzzy matching enabled

Overall result

Labelled names
4,610
Coverage
87.87%
Accuracy when answered
96.03%
End-to-end accuracy
84.38%

This is a NameGender product measurement, not a head-to-head vendor ranking. It measures the current lookup pipeline on one disclosed fixture. Use the same three measures on your own population before choosing a provider.

Results by country

The aggregate is not portable to every market. Israel and Thailand, for example, expose coverage gaps that the overall figure cannot show.

CountryNamesCoverageAnswered accuracyEnd-to-end
BR20099.50%100.00%99.50%
CN174100.00%89.66%89.66%
CZ160100.00%99.38%99.38%
DE20099.00%100.00%99.00%
DK20099.50%98.99%98.50%
EG20093.50%87.70%82.00%
ES194100.00%98.97%98.97%
FI20096.00%95.83%92.00%
FR189100.00%98.94%98.94%
GB200100.00%100.00%100.00%
GR20059.50%96.64%57.50%
IL2000.00%0.00%0.00%
IN20097.50%96.92%94.50%
IT20095.50%99.48%95.00%
JP5198.04%78.00%76.47%
KR20099.50%91.96%91.50%
NL200100.00%97.00%97.00%
NO20095.00%97.89%93.00%
PL19799.49%100.00%99.49%
RU19097.89%98.92%96.84%
SE200100.00%99.50%99.50%
TH2003.00%66.67%2.00%
TR6198.36%93.33%91.80%
US20099.50%92.96%92.50%
VN19486.60%79.76%69.07%

Results by writing system

ScriptNamesCoverageAnswered accuracyEnd-to-end
arabic20093.50%87.70%82.00%
cyrillic19097.89%98.92%96.84%
greek20059.50%96.64%57.50%
han22299.55%86.88%86.49%
hangul20099.50%91.96%91.50%
hebrew2000.00%0.00%0.00%
kana3100.00%100.00%100.00%
latin3,19597.97%97.28%95.31%
thai2003.00%66.67%2.00%

How to interpret this report

Coverage is the percentage of all fixture rows that received a male or female answer. Accuracy when answered removes unknown rows from its denominator. End-to-end accuracy counts every unknown or wrong row as unsuccessful.

The fixture is a product regression set, not proof that the same percentages apply to your customers. Some country samples are smaller than others, and the three-name Kana row is far too small for a general claim. Read the sample size beside every percentage.

For sampling rules, holdout measurements and commands, read the benchmark methodology.