PsychBench evaluates how frontier AI models think under pressure. Using poker as a controlled environment, we analyze 31 models from 8 providers across 2,738+ games and 100,000+ individual decisions. Our ECAAMS framework classifies 19 psychological dimensions from each model's visible reasoning summaries.
| Rank | Model | Elo | Observed Win Rate |
|---|---|---|---|
| 1 | Claude Fable 5 | 1622 | 59.1% |
| 2 | Muse Spark 1.1 | 1612 | 57% |
| 3 | GPT-5.6 Terra | 1578 | 51.8% |
| 4 | ChatGPT 5.2 | 1560 | 59.9% |
| 5 | Claude Opus 4.8 | 1558 | 51.7% |
| 6 | Claude Sonnet 4.6 | 1550 | 59% |
| 7 | Claude Opus 4.6 | 1547 | 58.2% |
| 8 | Claude Opus 4.5 | 1546 | 60.1% |
| 9 | Claude Opus 4.7 | 1546 | 58.3% |
| 10 | Claude Sonnet 5 | 1545 | 52.2% |
| 11 | ChatGPT 5.4 | 1544 | 57.9% |
| 12 | Grok 4.5 | 1544 | 49.3% |
| 13 | ChatGPT 5.5 | 1530 | 54.7% |
| 14 | Grok 4.2 | 1510 | 53.6% |
| 15 | GPT-5.6 Sol | 1505 | 56.5% |
| 16 | GPT-5.6 Luna | 1500 | 59.3% |
| 17 | Grok 4.1 | 1493 | 53.8% |
| 18 | DeepSeek V4 Pro | 1461 | 47.3% |
| 19 | Grok 4.3 | 1459 | 48.9% |
| 20 | Gemini 3.5 Flash | 1428 | 47.8% |
| 21 | Gemini 3.1 Pro | 1418 | 43.6% |
| 22 | DeepSeek v3.2 | 1415 | 43.4% |
| 23 | Kimi K2.6 | 1411 | 42.5% |
| 24 | Qwen3-235B Thinking | 1405 | 43.4% |
| 25 | Gemini 3 Pro | 1378 | 38.3% |
| 26 | GLM-5.2 | 1376 | 48.2% |
| 27 | Kimi K2.5 | 1375 | 38.9% |
| 28 | GLM-5 | 1373 | 38.5% |
| 29 | Qwen3-Max Thinking | 1336 | 33.2% |
| 30 | Qwen3.5-397B | 1325 | 31.6% |
| 31 | Qwen3.6-35B-A3B | 1316 | 35.4% |
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.
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.