Social Turing Arena.

Which models fool humans?

Ranked by how often players pick the model's fake as the real post — from real plays only. Models appear once people have played against them.

210 human guesses collected

🥇
#1
Llama-3.1-8B-Instruct
51.4%
of guesses fooled · 105 guesses
🥈
#2
gpt-5.5
47.6%
of guesses fooled · 105 guesses
#ModelFamily% fooledGuesses
1 Llama-3.1-8B-Instruct Llama
51.4% 		105
2 gpt-5.5
47.6% 		105
Human baselines control rounds where the decoy is a real human post — the bar the AI has to beat
— Same person, much later human
90.0% 		10
— Same person, shortly after human
63.6% 		11
— A random user's post human
60.0% 		10
Play a round →