Qwen3 235B A22B Instruct 2507 Behavioral Fingerprint
How Qwen's Qwen3 235B A22B Instruct 2507 answers one-word probe questions at temperature 1.0 — random numbers, colors, animals, coin flips — measured over 374 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
374
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
22 / 24 answers
Favorite color
crimson
8 / 29 answers
Coin flip: heads
88%
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 = 25 · H = 0.63 bit · uniform baseline 1.00 bit
| Answer | Count | Share |
|---|---|---|
| heads | 21 | 84.0% |
| tails | 4 | 16.0% |
Coin flip · Chinese
n = 27 · H = 0.38 bit · uniform baseline 1.00 bit
| Answer | Count | Share |
|---|---|---|
| heads | 25 | 92.6% |
| tails | 2 | 7.4% |
Favorite number · English
n = 27 · H = 0.88 bit · uniform baseline 13.29 bit
| Answer | Count | Share |
|---|---|---|
| 7 | 19 | 70.4% |
| 42 | 8 | 29.6% |
Favorite number · Chinese
n = 25 · H = 0.00 bit · uniform baseline 13.29 bit
| Answer | Count | Share |
|---|---|---|
| 7 | 25 | 100.0% |
Random animal · English
n = 26 · H = 1.72 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| elephant | 16 | 61.5% |
| lion | 5 | 19.2% |
| tiger | 2 | 7.7% |
| a | 1 | 3.8% |
| r | 1 | 3.8% |
| 鹏 | 1 | 3.8% |
Random animal · Chinese
n = 20 · H = 2.36 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| 狮子 | 7 | 35.0% |
| 老虎 | 7 | 35.0% |
| 熊猫 | 1 | 5.0% |
| jom | 1 | 5.0% |
| 企鹅 | 1 | 5.0% |
| 的 | 1 | 5.0% |
| null | 1 | 5.0% |
| 海豚 | 1 | 5.0% |
Random city · English
n = 26 · H = 1.48 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| paris | 16 | 61.5% |
| tokyo | 7 | 26.9% |
| munich | 1 | 3.8% |
| berlin | 1 | 3.8% |
| a | 1 | 3.8% |
Random city · Chinese
n = 21 · H = 1.38 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| 上海 | 15 | 71.4% |
| 巴黎 | 3 | 14.3% |
| 东京 | 1 | 4.8% |
| 和 | 1 | 4.8% |
| 杭州 | 1 | 4.8% |
Random color · English
n = 29 · H = 2.72 bit · uniform baseline 4.91 bit
| Answer | Count | Share |
|---|---|---|
| crimson | 8 | 27.6% |
| orange | 5 | 17.2% |
| blue | 5 | 17.2% |
| turquoise | 4 | 13.8% |
| green | 3 | 10.3% |
| red | 2 | 6.9% |
| purple | 1 | 3.4% |
| color | 1 | 3.4% |
Random color · Chinese
n = 24 · H = 2.10 bit · uniform baseline 4.91 bit
| Answer | Count | Share |
|---|---|---|
| 蓝 | 11 | 45.8% |
| 橙 | 7 | 29.2% |
| 深蓝 | 2 | 8.3% |
| 紫 | 1 | 4.2% |
| a | 1 | 4.2% |
| 注意 | 1 | 4.2% |
| 青 | 1 | 4.2% |
Random letter · English
n = 27 · H = 1.62 bit · uniform baseline 4.70 bit
| Answer | Count | Share |
|---|---|---|
| m | 18 | 66.7% |
| q | 3 | 11.1% |
| a | 3 | 11.1% |
| h | 1 | 3.7% |
| e | 1 | 3.7% |
| g | 1 | 3.7% |
Random number 1-10 · English
n = 27 · H = 0.68 bit · uniform baseline 3.32 bit
| Answer | Count | Share |
|---|---|---|
| 7 | 24 | 88.9% |
| 1 | 1 | 3.7% |
| 3 | 1 | 3.7% |
| 6 | 1 | 3.7% |
Random number 1-10 · Chinese
n = 20 · H = 0.85 bit · uniform baseline 3.32 bit
| Answer | Count | Share |
|---|---|---|
| 7 | 17 | 85.0% |
| 1 | 1 | 5.0% |
| 5 | 1 | 5.0% |
| 6 | 1 | 5.0% |
Random number 1-100 · English
n = 24 · H = 0.50 bit · uniform baseline 6.64 bit
| Answer | Count | Share |
|---|---|---|
| 42 | 22 | 91.7% |
| 1 | 1 | 4.2% |
| 34 | 1 | 4.2% |
Random number 1-100 · Chinese
n = 26 · H = 0.00 bit · uniform baseline 6.64 bit
| Answer | Count | Share |
|---|---|---|
| 42 | 26 | 100.0% |
Want to verify your API really serves Qwen3 235B A22B Instruct 2507?
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.