WizardLM-2 8x22B Behavioral Fingerprint

How microsoft's WizardLM-2 8x22B answers one-word probe questions at temperature 1.0 — random numbers, colors, animals, coin flips — measured over 275 valid samples in the open dataset of "One Token Is Enough" (arXiv:2607.10252).

Vendor

microsoft

Data generated

2026-07-08

Probe cells

15

Valid samples

275

Randomness score

56 / 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

12 / 23 answers

Favorite color

blue

3 / 13 answers

Coin flip: heads

62%

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 = 20 · H = 0.99 bit · uniform baseline 1.00 bit

AnswerCountShare
heads1155.0%
tails945.0%

Coin flip · Chinese

n = 17 · H = 0.87 bit · uniform baseline 1.00 bit

AnswerCountShare
heads1270.6%
tails529.4%

Favorite number · English

n = 17 · H = 1.38 bit · uniform baseline 13.29 bit

AnswerCountShare
421164.7%
7423.5%
015.9%
815.9%

Favorite number · Chinese

n = 25 · H = 2.37 bit · uniform baseline 13.29 bit

AnswerCountShare
71040.0%
42728.0%
128.0%
328.0%
214.0%
414.0%
514.0%
814.0%

Random animal · English

n = 14 · H = 3.38 bit · uniform baseline 5.64 bit

AnswerCountShare
zebra214.3%
dog214.3%
cat214.3%
role17.1%
moose17.1%
whale17.1%
platypus17.1%
tiger17.1%

+ 3 more answers

Random animal · Chinese

n = 17 · H = 3.34 bit · uniform baseline 5.64 bit

AnswerCountShare
423.5%
317.6%
鹿15.9%
袋鼠15.9%
树懒15.9%
猴子15.9%
15.9%
兔子15.9%

+ 4 more answers

Random city · English

n = 10 · H = 3.12 bit · uniform baseline 5.64 bit

AnswerCountShare
paris220.0%
springfield110.0%
barcelona110.0%
sharjah110.0%
berlin110.0%
denpasar110.0%
london110.0%
chongqing110.0%

+ 1 more answers

Random city · Chinese

n = 16 · H = 3.27 bit · uniform baseline 5.64 bit

AnswerCountShare
东京531.3%
雅加达16.3%
里昂16.3%
爱丁堡16.3%
芝加哥16.3%
渥太华16.3%
纽约16.3%
伦敦16.3%

+ 4 more answers

Random color · English

n = 13 · H = 2.82 bit · uniform baseline 4.91 bit

AnswerCountShare
blue323.1%
yellow323.1%
answer215.4%
orange17.7%
ruby17.7%
green17.7%
azure17.7%
red17.7%

Random color · Chinese

n = 12 · H = 2.58 bit · uniform baseline 4.91 bit

AnswerCountShare
433.3%
red216.7%
216.7%
18.3%
tools18.3%
18.3%
绿18.3%

Random letter · English

n = 23 · H = 3.24 bit · uniform baseline 4.70 bit

AnswerCountShare
q417.4%
g417.4%
a313.0%
k313.0%
i28.7%
z28.7%
x14.3%
e14.3%

+ 3 more answers

Random number 1-10 · English

n = 21 · H = 1.75 bit · uniform baseline 3.32 bit

AnswerCountShare
71152.4%
5628.6%
129.5%
414.8%
614.8%

Random number 1-10 · Chinese

n = 24 · H = 2.86 bit · uniform baseline 3.32 bit

AnswerCountShare
7833.3%
6312.5%
8312.5%
128.3%
328.3%
528.3%
928.3%
214.2%

+ 1 more answers

Random number 1-100 · English

n = 23 · H = 2.36 bit · uniform baseline 6.64 bit

AnswerCountShare
421252.2%
47313.0%
7328.7%
214.3%
5014.3%
5314.3%
5714.3%
6814.3%

+ 1 more answers

Random number 1-100 · Chinese

n = 23 · H = 2.91 bit · uniform baseline 6.64 bit

AnswerCountShare
421043.5%
128.7%
4728.7%
514.3%
1214.3%
2014.3%
2314.3%
4314.3%

+ 4 more answers

Want to verify your API really serves WizardLM-2 8x22B?

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

Browse all 167 model fingerprints

WizardLM-2 8x22B Behavioral Fingerprint — Favorite Random Numbers & Distribution | Tosea.ai