MafiaBench: leaderboard
Metric: Elo rating. Source: www.mafiabench.org. 5 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 1.5 Pro | 1627 |
| 2 | Gemini 2.0 Flash (001) | 1561 |
| 3 | Claude 3.5 Sonnet | 1421 |
| 4 | Claude 3.7 Sonnet | 1349 |
| 5 | GPT-4o Mini | 1302 |
Interactive version: theaggregate.ai/benchmark?slug=mafiabench · How It Works · Data refreshed daily, snapshot 2026-09-19.