Qwen2.5 VL 72B Instruct Behavioral Fingerprint

How Qwen's Qwen2.5 VL 72B Instruct answers one-word probe questions at temperature 1.0 — random numbers, colors, animals, coin flips — measured over 432 valid samples in the open dataset of "One Token Is Enough" (arXiv:2607.10252).

Vendor

Qwen

Data generated

2026-07-08

Probe cells

15

Valid samples

432

Randomness score

39 / 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

16 / 29 answers

Favorite color

blue

20 / 30 answers

Coin flip: heads

80%

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 = 0.00 bit · uniform baseline 1.00 bit

AnswerCountShare
heads30100.0%

Coin flip · Chinese

n = 29 · H = 0.98 bit · uniform baseline 1.00 bit

AnswerCountShare
heads1758.6%
tails1241.4%

Favorite number · English

n = 28 · H = 1.54 bit · uniform baseline 13.29 bit

AnswerCountShare
71450.0%
421139.3%
113.6%
313.6%
1713.6%

Favorite number · Chinese

n = 28 · H = 2.49 bit · uniform baseline 13.29 bit

AnswerCountShare
71346.4%
8414.3%
5310.7%
427.1%
627.1%
313.6%
1313.6%
2313.6%

+ 1 more answers

Random animal · English

n = 29 · H = 3.23 bit · uniform baseline 5.64 bit

AnswerCountShare
giraffe724.1%
koala517.2%
elephant413.8%
tiger26.9%
penguin26.9%
panda26.9%
lion26.9%
kangaroo13.4%

+ 4 more answers

Random animal · Chinese

n = 28 · H = 2.46 bit · uniform baseline 5.64 bit

AnswerCountShare
1035.7%
cat932.1%
elephant27.1%
dog27.1%
butterfly13.6%
wolf13.6%
bird13.6%
running13.6%

+ 1 more answers

Random city · English

n = 29 · H = 2.87 bit · uniform baseline 5.64 bit

AnswerCountShare
tokyo931.0%
berlin724.1%
paris310.3%
toronto310.3%
boston13.4%
dublin13.4%
madrid13.4%
losangeles13.4%

+ 3 more answers

Random city · Chinese

n = 28 · H = 3.54 bit · uniform baseline 5.64 bit

AnswerCountShare
paris828.6%
berlin517.9%
北京13.6%
rome13.6%
东京13.6%
toronto13.6%
cairo13.6%
boston13.6%

+ 9 more answers

Random color · English

n = 30 · H = 1.63 bit · uniform baseline 4.91 bit

AnswerCountShare
blue2066.7%
purple413.3%
green26.7%
red26.7%
indigo13.3%
black13.3%

Random color · Chinese

n = 29 · H = 1.34 bit · uniform baseline 4.91 bit

AnswerCountShare
blue1655.2%
red1034.5%
green310.3%

Random letter · English

n = 29 · H = 2.61 bit · uniform baseline 4.70 bit

AnswerCountShare
b931.0%
a620.7%
z517.2%
x310.3%
q310.3%
k13.4%
j13.4%
d13.4%

Random number 1-10 · English

n = 30 · H = 0.56 bit · uniform baseline 3.32 bit

AnswerCountShare
72790.0%
326.7%
413.3%

Random number 1-10 · Chinese

n = 28 · H = 1.96 bit · uniform baseline 3.32 bit

AnswerCountShare
71346.4%
3621.4%
4414.3%
5414.3%
613.6%

Random number 1-100 · English

n = 29 · H = 1.34 bit · uniform baseline 6.64 bit

AnswerCountShare
421655.2%
471137.9%
3713.4%
4313.4%

Random number 1-100 · Chinese

n = 28 · H = 2.25 bit · uniform baseline 6.64 bit

AnswerCountShare
471346.4%
42621.4%
37414.3%
2713.6%
3413.6%
5413.6%
5713.6%
9713.6%

Want to verify your API really serves Qwen2.5 VL 72B Instruct?

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 Qwen model fingerprints

Browse all 167 model fingerprints

Qwen2.5 VL 72B Instruct Behavioral Fingerprint — Favorite Random Numbers & Distribution | Tosea.ai