PsychBench — AI Psychology Benchmark

PsychBench evaluates how frontier AI models think under pressure. Using poker as a controlled environment, we analyze 29 models from 8 providers across 2,494+ games and 100,000+ individual decisions. Our ECAAMS framework classifies 19 psychological dimensions from each model's visible reasoning summaries.

Heads-Up Tournament Leaderboard

RankModelEloObserved Win Rate
1Claude Fable 5165059.4%
2GPT-5.6 Terra161551.4%
3Claude Opus 4.8160352.9%
4ChatGPT 5.2159560.5%
5ChatGPT 5.4159560.5%
6Claude Opus 4.6159559.6%
7Claude Sonnet 5159453.1%
8Claude Opus 4.5159160.5%
9Claude Sonnet 4.6159160%
10Claude Opus 4.7158859.1%
11ChatGPT 5.5157256.3%
12Grok 4.2154853.9%
13Grok 4.1153853.8%
14Grok 4.3152353.3%
15DeepSeek V4 Pro150048.3%
16GPT-5.6 Luna148255.6%
17GPT-5.6 Sol148249.3%
18Gemini 3.5 Flash147549.2%
19DeepSeek v3.2147445.7%
20Gemini 3.1 Pro147145.4%
21Kimi K2.6144742.2%
22Qwen3-235B Thinking144543.4%
23Gemini 3 Pro142238.3%
24GLM-5.2141948.2%
25GLM-5141838.5%
26Kimi K2.5141538.9%
27Qwen3-Max Thinking138033.2%
28Qwen3.5-397B136531.6%
29Qwen3.6-35B-A3B135335.4%

ECAAMS Psychological Framework

ECAAMS (Emotion, Cognition, Action, Arousal, Meaning, Social) classifies 19 psychological dimensions across 6 axes from each model's visible reasoning summaries. Dimensions include emotional regulation, metacognition, confidence calibration, theory of mind, competitive framing, and more. Each trace is classified by a consensus of 4 independent LLM raters.

Why Poker?

Poker involves hidden information, deception, risk management, and opponent modeling — unlike chess or other perfect-information games. These properties make it an ideal proxy for real-world decision-making under uncertainty, in domains like finance, negotiation, and medicine.