BehaviorBench - Cross-Game Prediction: leaderboard
Metric: Given the first-round moves of one subject in other games, predict the first-round move in a target game, on MobLab economic-game records of human players (dictator, ultimatum, trust, public goods, bomb risk, beauty contest and push/pull games): Mean absolute error of the predicted move in the game choice units; lower is better. Source: arxiv.org. Saturation forecast: Around August 2027. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 | 25.3 |
| 2 | Claude Opus 4.6 | 25.9 |
| 3 | Claude Haiku 4.5 | 26 |
| 4 | GPT-5.4 (High) | 26.1 |
| 5 | Llama 3.3 70B Instruct | 26.4 |
| 6 | Gemini 3.1 Pro (Preview) | 26.7 |
| 7 | GPT-4.1 | 27.1 |
| 8 | GPT-5.4 Mini (High) | 27.4 |
| 9 | DeepSeek V3.2 | 27.5 |
| 10 | Qwen 3 4B | 27.9 |
| 11 | Gemini 3.1 Flash Lite (Preview) | 29.4 |
Interactive version: theaggregate.ai/benchmark?slug=behaviorbench-cross-game-prediction · How It Works · Data refreshed daily, snapshot 2026-09-29.