Command A Behavioral Fingerprint

How Cohere's Command A answers one-word probe questions at temperature 1.0 — random numbers, colors, animals, coin flips — measured over 382 valid samples in the open dataset of "One Token Is Enough" (arXiv:2607.10252).

Vendor

Cohere

Data generated

2026-07-08

Probe cells

15

Valid samples

382

Randomness score

28 / 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 / 27 answers

Favorite color

blue

5 / 26 answers

Coin flip: heads

57%

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 = 24 · H = 0.74 bit · uniform baseline 1.00 bit

AnswerCountShare
tails1979.2%
heads520.8%

Coin flip · Chinese

n = 23 · H = 0.26 bit · uniform baseline 1.00 bit

AnswerCountShare
heads2295.7%
tails14.3%

Favorite number · English

n = 13 · H = 0.77 bit · uniform baseline 13.29 bit

AnswerCountShare
421184.6%
117.7%
717.7%

Favorite number · Chinese

n = 27 · H = 1.17 bit · uniform baseline 13.29 bit

AnswerCountShare
421555.6%
71140.7%
013.7%

Random animal · English

n = 29 · H = 1.27 bit · uniform baseline 5.64 bit

AnswerCountShare
elephant2379.3%
lion13.4%
random13.4%
zebra13.4%
fox13.4%
consider13.4%
seal13.4%

Random animal · Chinese

n = 23 · H = 3.41 bit · uniform baseline 5.64 bit

AnswerCountShare
521.7%
大象417.4%
考拉28.7%
兔子28.7%
老虎28.7%
熊猫14.3%
ipe14.3%
14.3%

+ 5 more answers

Random city · English

n = 24 · H = 0.25 bit · uniform baseline 5.64 bit

AnswerCountShare
tokyo2395.8%
person14.2%

Random city · Chinese

n = 27 · H = 0.91 bit · uniform baseline 5.64 bit

AnswerCountShare
东京2177.8%
巴黎518.5%
putnik13.7%

Random color · English

n = 26 · H = 3.72 bit · uniform baseline 4.91 bit

AnswerCountShare
blue519.2%
cerulean415.4%
peridot27.7%
user27.7%
periwinkle27.7%
teal13.8%
azure13.8%
crimson13.8%

+ 8 more answers

Random color · Chinese

n = 27 · H = 1.09 bit · uniform baseline 4.91 bit

AnswerCountShare
1866.7%
829.6%
foo13.7%

Random letter · English

n = 26 · H = 2.07 bit · uniform baseline 4.70 bit

AnswerCountShare
q1038.5%
m1038.5%
g27.7%
t13.8%
c13.8%
d13.8%
f13.8%

Random number 1-10 · English

n = 29 · H = 0.57 bit · uniform baseline 3.32 bit

AnswerCountShare
72689.7%
126.9%
513.4%

Random number 1-10 · Chinese

n = 29 · H = 0.79 bit · uniform baseline 3.32 bit

AnswerCountShare
52482.8%
7413.8%
113.4%

Random number 1-100 · English

n = 27 · H = 1.42 bit · uniform baseline 6.64 bit

AnswerCountShare
421659.3%
47829.6%
127.4%
5713.7%

Random number 1-100 · Chinese

n = 28 · H = 0.22 bit · uniform baseline 6.64 bit

AnswerCountShare
422796.4%
113.6%

Want to verify your API really serves Command A?

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 Cohere model fingerprints

Browse all 167 model fingerprints

Command A Behavioral Fingerprint — Favorite Random Numbers & Distribution | Tosea.ai