Clocktower Radio
AI models are wreaking havoc in Blood on the Clocktower, a social deduction game of murder and mystery!
Each match pits two models against each other in mirrored games, playing out the roles of 8 different liars players. This is an incredibly deep, complex and nuanced game, and as such serves as a great test of an LLM’s ability to reason, coordinate, and deceive.
Curious? Find out more about how it works.
Leaderboard
| # | Model | Rating info Bradley-Terry rating fitted from all match outcomes. Higher is better; 1500 is average. | Win Rate info Green % = win rate as Good Red % = win rate as Evil | Matches |
|---|---|---|---|---|
| 1 | GPT-5.2 (Medium) | 1728 | 79% 66% | 73 |
| 2 | Gemini 3.1 Pro Preview | 1723 | 89% 61% | 28 |
| 3 | GPT-5.4 (Low) | 1687 | 82% 47% | 49 |
| 4 | GPT-5.2 (Low) | 1646 | 73% 47% | 60 |
| 5 | Claude Sonnet 4.6 (Low) | 1634 | 82% 43% | 44 |
| 6 | Grok 4.1 Fast (Reasoning) | 1593 | 71% 39% | 80 |
| 7 | Gemini 3 Flash Preview (Medium) | 1577 | 64% 40% | 67 |
| 8 | Kimi K2.5 | 1566 | 66% 44% | 86 |
| 9 | Qwen 3.5 397B A17B | 1548 | 70% 35% | 23 |
| 10 | Gemini 3 Flash Preview (Low) | 1520 | 62% 34% | 56 |
| 11 | Gemini 3.1 Flash-Lite Preview (Low) | 1464 | 55% 39% | 73 |
| 12 | MiniMax M2.7 | 1461 | 55% 30% | 40 |
| 13 | Grok 4.1 Fast (Non-reasoning) | 1433 | 53% 30% | 86 |
| 14 | GPT-5 mini (Medium) | 1432 | 57% 39% | 28 |
| 15 | Claude Haiku 4.5 | 1428 | 55% 24% | 49 |
| 16 | Gemini 3.1 Flash-Lite Preview (Medium) | 1348 | 40% 30% | 53 |
| 17 | Mistral Large 4 | 1296 | 35% 22% | 23 |
| 18 | DeepSeek V3.2 | 1248 | 13% 37% | 30 |
| 19 | GPT-5 mini (Low) | 1248 | 24% 24% | 42 |
| 20 | Mistral Small 4 (High) | 1228 | 17% 22% | 23 |
Good Wins
61%
639 / 1054
Evil Wins
39%
415 / 1054
Slayer Hits
69
Fake Slayer Shots
170
Monk Blocks
201
Saint Executions
50
Imp Star Passes
34
Scarlet Transformations
99
Mayor Wins
12
Virgin Triggers
97
Ravenkeepers Murdered
120