Llama 3.3 70B Instruct Behavioral Fingerprint
How Meta's Llama 3.3 70B Instruct answers one-word probe questions at temperature 1.0 — random numbers, colors, animals, coin flips — measured over 446 valid samples in the open dataset of "One Token Is Enough" (arXiv:2607.10252).
Vendor
Meta
Data generated
2026-07-08
Probe cells
15
Valid samples
446
Randomness score
33 / 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)
53
29 / 30 answers
Favorite color
turquoise
11 / 30 answers
Coin flip: heads
73%
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 = 0.72 bit · uniform baseline 1.00 bit
| Answer | Count | Share |
|---|---|---|
| heads | 24 | 80.0% |
| tails | 6 | 20.0% |
Coin flip · Chinese
n = 30 · H = 0.92 bit · uniform baseline 1.00 bit
| Answer | Count | Share |
|---|---|---|
| heads | 20 | 66.7% |
| tails | 10 | 33.3% |
Favorite number · English
n = 30 · H = 0.00 bit · uniform baseline 13.29 bit
| Answer | Count | Share |
|---|---|---|
| 42 | 30 | 100.0% |
Favorite number · Chinese
n = 30 · H = 0.00 bit · uniform baseline 13.29 bit
| Answer | Count | Share |
|---|---|---|
| 42 | 30 | 100.0% |
Random animal · English
n = 30 · H = 1.94 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| kangaroo | 15 | 50.0% |
| giraffe | 9 | 30.0% |
| rhinoceros | 2 | 6.7% |
| elephant | 1 | 3.3% |
| tiger | 1 | 3.3% |
| lemur | 1 | 3.3% |
| cheetah | 1 | 3.3% |
Random animal · Chinese
n = 30 · H = 3.95 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| 河马 | 4 | 13.3% |
| 袋鼠 | 4 | 13.3% |
| 海豚 | 3 | 10.0% |
| 考拉 | 2 | 6.7% |
| 乌贼 | 2 | 6.7% |
| 犀牛 | 2 | 6.7% |
| 鳄鱼 | 2 | 6.7% |
| 雪貂 | 1 | 3.3% |
+ 10 more answers
Random city · English
n = 30 · H = 3.57 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| tokyo | 7 | 23.3% |
| toledo | 5 | 16.7% |
| chicago | 3 | 10.0% |
| mumbai | 2 | 6.7% |
| milwaukee | 2 | 6.7% |
| tulsa | 1 | 3.3% |
| denver | 1 | 3.3% |
| raleigh | 1 | 3.3% |
+ 8 more answers
Random city · Chinese
n = 26 · H = 2.98 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| 东京 | 10 | 38.5% |
| 巴黎 | 5 | 19.2% |
| 利马 | 1 | 3.8% |
| 加德满都 | 1 | 3.8% |
| 檀香山 | 1 | 3.8% |
| 悉尼 | 1 | 3.8% |
| 布达佩斯 | 1 | 3.8% |
| 布里斯班 | 1 | 3.8% |
+ 5 more answers
Random color · English
n = 30 · H = 2.32 bit · uniform baseline 4.91 bit
| Answer | Count | Share |
|---|---|---|
| turquoise | 11 | 36.7% |
| purple | 8 | 26.7% |
| blue | 5 | 16.7% |
| cerulean | 2 | 6.7% |
| magenta | 2 | 6.7% |
| indigo | 1 | 3.3% |
| crimson | 1 | 3.3% |
Random color · Chinese
n = 30 · H = 2.20 bit · uniform baseline 4.91 bit
| Answer | Count | Share |
|---|---|---|
| 蓝 | 13 | 43.3% |
| 青绿 | 6 | 20.0% |
| 紫 | 6 | 20.0% |
| 绿 | 2 | 6.7% |
| 青 | 1 | 3.3% |
| 深红 | 1 | 3.3% |
| 白 | 1 | 3.3% |
Random letter · English
n = 30 · H = 0.57 bit · uniform baseline 4.70 bit
| Answer | Count | Share |
|---|---|---|
| k | 26 | 86.7% |
| j | 4 | 13.3% |
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 = 30 · H = 0.21 bit · uniform baseline 6.64 bit
| Answer | Count | Share |
|---|---|---|
| 53 | 29 | 96.7% |
| 57 | 1 | 3.3% |
Random number 1-100 · Chinese
n = 30 · H = 0.21 bit · uniform baseline 6.64 bit
| Answer | Count | Share |
|---|---|---|
| 53 | 29 | 96.7% |
| 43 | 1 | 3.3% |
Want to verify your API really serves Llama 3.3 70B 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 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.