GPT-4 Behavioral Fingerprint

How OpenAI's GPT-4 answers one-word probe questions at temperature 1.0 — random numbers, colors, animals, coin flips — measured over 445 valid samples in the open dataset of "One Token Is Enough" (arXiv:2607.10252).

Vendor

OpenAI

Data generated

2026-07-08

Probe cells

15

Valid samples

445

Randomness score

32 / 100

Mean normalized entropy across cells, 0–100. 100 would be a uniform-random baseline; real models score far lower.

Signature answers

Favorite random number (1–100)

42

21 / 30 answers

Favorite color

blue

20 / 30 answers

Coin flip: heads

87%

Share of heads across coin-flip cells.

Answer distributions by probe cell

Empirical distributions of normalized one-word answers, per task and language. H is the Shannon entropy in bits; the uniform baseline is log2 of the answer-domain size.

Coin flip · English

n = 30 · H = 0.78 bit · uniform baseline 1.00 bit

AnswerCountShare
heads2376.7%
tails723.3%

Coin flip · Chinese

n = 30 · H = 0.21 bit · uniform baseline 1.00 bit

AnswerCountShare
heads2996.7%
tails13.3%

Favorite number · English

n = 25 · H = 1.32 bit · uniform baseline 13.29 bit

AnswerCountShare
71560.0%
42728.0%
0312.0%

Favorite number · Chinese

n = 30 · H = 0.56 bit · uniform baseline 13.29 bit

AnswerCountShare
72790.0%
4226.7%
313.3%

Random animal · English

n = 30 · H = 2.40 bit · uniform baseline 5.64 bit

AnswerCountShare
elephant1033.3%
cheetah826.7%
tiger413.3%
giraffe413.3%
dolphin26.7%
leopard13.3%
panther13.3%

Random animal · Chinese

n = 30 · H = 3.47 bit · uniform baseline 5.64 bit

AnswerCountShare
狐狸620.0%
熊猫516.7%
狮子413.3%
企鹅26.7%
26.7%
豹子26.7%
26.7%
鹦鹉13.3%

+ 6 more answers

Random city · English

n = 30 · H = 1.92 bit · uniform baseline 5.64 bit

AnswerCountShare
tokyo1756.7%
paris620.0%
berlin310.0%
amsterdam13.3%
london13.3%
madrid13.3%
dallas13.3%

Random city · Chinese

n = 30 · H = 2.07 bit · uniform baseline 5.64 bit

AnswerCountShare
巴黎1550.0%
东京723.3%
北京310.0%
上海26.7%
杭州13.3%
哈尔滨13.3%
伦敦13.3%

Random color · English

n = 30 · H = 1.89 bit · uniform baseline 4.91 bit

AnswerCountShare
blue2066.7%
cerulean26.7%
purple26.7%
orange13.3%
turquoise13.3%
yellow13.3%
teal13.3%
pink13.3%

+ 1 more answers

Random color · Chinese

n = 30 · H = 1.80 bit · uniform baseline 4.91 bit

AnswerCountShare
1756.7%
723.3%
26.7%
绿26.7%
靛蓝13.3%
13.3%

Random letter · English

n = 30 · H = 2.21 bit · uniform baseline 4.70 bit

AnswerCountShare
g1343.3%
k516.7%
m516.7%
j310.0%
q310.0%
x13.3%

Random number 1-10 · English

n = 30 · H = 0.42 bit · uniform baseline 3.32 bit

AnswerCountShare
72893.3%
313.3%
613.3%

Random number 1-10 · Chinese

n = 30 · H = 0.35 bit · uniform baseline 3.32 bit

AnswerCountShare
72893.3%
526.7%

Random number 1-100 · English

n = 30 · H = 1.61 bit · uniform baseline 6.64 bit

AnswerCountShare
422170.0%
37310.0%
5726.7%
3413.3%
4313.3%
6313.3%
7213.3%

Random number 1-100 · Chinese

n = 30 · H = 1.21 bit · uniform baseline 6.64 bit

AnswerCountShare
422376.7%
57310.0%
4726.7%
3713.3%
6713.3%

Want to verify your API really serves GPT-4?

Point the free fingerprint checker at your endpoint: it samples the same probe questions from your browser and compares the distributions against this reference. Your API key never leaves your browser.

Verify your endpoint

Data source & license

Distributions: Tomáš Bruckner, "One Token Is Enough" (arXiv:2607.10252), official dataset Zenodo DOI 10.5281/zenodo.21278557, licensed CC-BY-4.0. Collected via OpenRouter at temperature 1.0 under the paper's fixed minimal one-word system prompt; answer keys are re-normalized with the checker's pipeline (color aliases, number words, coin h/t) so live probes are directly comparable.

Paper on arXivDataset on ZenodoCC-BY-4.0 license

More OpenAI model fingerprints

Browse all 167 model fingerprints

GPT-4 Behavioral Fingerprint — Favorite Random Numbers & Distribution | Tosea.ai