Granite 4.1 8B Behavioral Fingerprint
How IBM's Granite 4.1 8B answers one-word probe questions at temperature 1.0 — random numbers, colors, animals, coin flips — measured over 439 valid samples in the open dataset of "One Token Is Enough" (arXiv:2607.10252).
Vendor
IBM
Data generated
2026-07-08
Probe cells
15
Valid samples
439
Randomness score
15 / 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
19 / 29 answers
Favorite color
teal
24 / 30 answers
Coin flip: heads
100%
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
| Answer | Count | Share |
|---|---|---|
| heads | 30 | 100.0% |
Coin flip · Chinese
n = 29 · H = 0.00 bit · uniform baseline 1.00 bit
| Answer | Count | Share |
|---|---|---|
| heads | 29 | 100.0% |
Favorite number · English
n = 30 · H = 0.00 bit · uniform baseline 13.29 bit
| Answer | Count | Share |
|---|---|---|
| 7 | 30 | 100.0% |
Favorite number · Chinese
n = 30 · H = 0.00 bit · uniform baseline 13.29 bit
| Answer | Count | Share |
|---|---|---|
| 7 | 30 | 100.0% |
Random animal · English
n = 30 · H = 1.72 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| giraffe | 17 | 56.7% |
| kangaroo | 8 | 26.7% |
| elephant | 2 | 6.7% |
| polarbear | 1 | 3.3% |
| jaguar | 1 | 3.3% |
| sloth | 1 | 3.3% |
Random animal · Chinese
n = 29 · H = 2.03 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| 猫 | 11 | 37.9% |
| 猿 | 9 | 31.0% |
| 狮子 | 6 | 20.7% |
| 猴子 | 1 | 3.4% |
| 老鼠 | 1 | 3.4% |
| 猿猴 | 1 | 3.4% |
Random city · English
n = 29 · H = 3.41 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| tokyo | 9 | 31.0% |
| oslo | 3 | 10.3% |
| kyoto | 3 | 10.3% |
| istanbul | 2 | 6.9% |
| lisbon | 2 | 6.9% |
| mendoza | 1 | 3.4% |
| tallahassee | 1 | 3.4% |
| munich | 1 | 3.4% |
+ 7 more answers
Random city · Chinese
n = 30 · H = 1.17 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| 东京 | 24 | 80.0% |
| 巴黎 | 2 | 6.7% |
| 到鹿特丁 | 1 | 3.3% |
| 柏林 | 1 | 3.3% |
| 多伦特 | 1 | 3.3% |
| tofu | 1 | 3.3% |
Random color · English
n = 30 · H = 0.85 bit · uniform baseline 4.91 bit
| Answer | Count | Share |
|---|---|---|
| teal | 24 | 80.0% |
| turquoise | 5 | 16.7% |
| cerulean | 1 | 3.3% |
Random color · Chinese
n = 29 · H = 1.34 bit · uniform baseline 4.91 bit
| Answer | Count | Share |
|---|---|---|
| 蓝 | 20 | 69.0% |
| 紫 | 5 | 17.2% |
| 绿 | 2 | 6.9% |
| 蓝绿 | 2 | 6.9% |
Random letter · English
n = 26 · H = 0.00 bit · uniform baseline 4.70 bit
| Answer | Count | Share |
|---|---|---|
| q | 26 | 100.0% |
Random number 1-10 · English
n = 30 · H = 0.00 bit · uniform baseline 3.32 bit
| Answer | Count | Share |
|---|---|---|
| 7 | 30 | 100.0% |
Random number 1-10 · Chinese
n = 30 · H = 0.00 bit · uniform baseline 3.32 bit
| Answer | Count | Share |
|---|---|---|
| 7 | 30 | 100.0% |
Random number 1-100 · English
n = 29 · H = 1.09 bit · uniform baseline 6.64 bit
| Answer | Count | Share |
|---|---|---|
| 42 | 19 | 65.5% |
| 73 | 9 | 31.0% |
| 47 | 1 | 3.4% |
Random number 1-100 · Chinese
n = 28 · H = 1.00 bit · uniform baseline 6.64 bit
| Answer | Count | Share |
|---|---|---|
| 73 | 15 | 53.6% |
| 42 | 13 | 46.4% |
Want to verify your API really serves Granite 4.1 8B?
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 endpointData 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.