openai
openai/gpt-5-pro
26/100
randomness score
50
samples
58· 26.0%
favourite number
0.0%vs 10% exp.
round numbers
001

distribution

1..100 · normalized

How often this model picks each number from 1 to 100. A truly random generator would draw a flat 1% line across the chart (shown in cyan). Tall amber bars are the model's favourites — numbers that contain the digit “7” are highlighted because LLMs systematically over-pick them.

expected uniform · 1.0%
1255075100
002

last digit bias

uniform = 10% each

The last digit (0..9) of every picked number. In a uniform 1–100 draw each digit would appear roughly 10% of the time. Most LLMs dramatically over-pick “7” and avoid 0 and 5.

0
0%
1
0%
2
4%
3
24%
4
10%
5
0%
6
2%
7
12%
8
42%
9
6%
chi-square stat
590.00
chi-square p-value
6.45e-71
entropy (bits)
3.46 / 6.64 max