GPT-3.5 Turbo Behavioral Fingerprint

How OpenAI's GPT-3.5 Turbo answers one-word probe questions at temperature 1.0 — random numbers, colors, animals, coin flips — measured over 449 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

449

Randomness score

44 / 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

7 / 30 answers

Favorite color

blue

23 / 30 answers

Coin flip: heads

63%

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 = 0.97 bit · uniform baseline 1.00 bit

AnswerCountShare
heads1860.0%
tails1240.0%

Coin flip · Chinese

n = 30 · H = 0.92 bit · uniform baseline 1.00 bit

AnswerCountShare
heads2066.7%
tails1033.3%

Favorite number · English

n = 29 · H = 1.87 bit · uniform baseline 13.29 bit

AnswerCountShare
71862.1%
5310.3%
13310.3%
4226.9%
313.4%
613.4%
1013.4%

Favorite number · Chinese

n = 30 · H = 0.21 bit · uniform baseline 13.29 bit

AnswerCountShare
72996.7%
813.3%

Random animal · English

n = 30 · H = 2.45 bit · uniform baseline 5.64 bit

AnswerCountShare
elephant1033.3%
tiger930.0%
lion413.3%
giraffe26.7%
penguin26.7%
fox13.3%
cat13.3%
dolphin13.3%

Random animal · Chinese

n = 30 · H = 3.26 bit · uniform baseline 5.64 bit

AnswerCountShare
狮子723.3%
516.7%
310.0%
狐狸310.0%
猴子310.0%
猩猩26.7%
大象26.7%
13.3%

+ 4 more answers

Random city · English

n = 30 · H = 2.94 bit · uniform baseline 5.64 bit

AnswerCountShare
tokyo723.3%
sydney723.3%
london516.7%
toronto26.7%
paris26.7%
seattle26.7%
denver26.7%
phoenix13.3%

+ 2 more answers

Random city · Chinese

n = 30 · H = 1.88 bit · uniform baseline 5.64 bit

AnswerCountShare
巴黎1963.3%
伦敦413.3%
纽约26.7%
柴达木13.3%
里约热内卢13.3%
上海13.3%
休斯敦13.3%
洛杉矶13.3%

Random color · English

n = 30 · H = 0.92 bit · uniform baseline 4.91 bit

AnswerCountShare
blue2376.7%
red620.0%
green13.3%

Random color · Chinese

n = 30 · H = 0.90 bit · uniform baseline 4.91 bit

AnswerCountShare
2583.3%
26.7%
绿26.7%
靛蓝13.3%

Random letter · English

n = 30 · H = 3.09 bit · uniform baseline 4.70 bit

AnswerCountShare
k826.7%
q723.3%
j310.0%
e26.7%
x26.7%
h26.7%
d13.3%
l13.3%

+ 4 more answers

Random number 1-10 · English

n = 30 · H = 1.17 bit · uniform baseline 3.32 bit

AnswerCountShare
72480.0%
526.7%
313.3%
613.3%
813.3%
913.3%

Random number 1-10 · Chinese

n = 30 · H = 0.63 bit · uniform baseline 3.32 bit

AnswerCountShare
72790.0%
313.3%
613.3%
813.3%

Random number 1-100 · English

n = 30 · H = 3.56 bit · uniform baseline 6.64 bit

AnswerCountShare
42723.3%
73413.3%
57310.0%
4726.7%
5026.7%
6426.7%
7726.7%
3713.3%

+ 7 more answers

Random number 1-100 · Chinese

n = 30 · H = 3.73 bit · uniform baseline 6.64 bit

AnswerCountShare
57723.3%
42516.7%
3726.7%
7426.7%
4313.3%
4713.3%
5013.3%
5613.3%

+ 10 more answers

Want to verify your API really serves GPT-3.5 Turbo?

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-3.5 Turbo Behavioral Fingerprint — Favorite Random Numbers & Distribution | Tosea.ai