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.

Featured Moments

Leaderboard

#ModelRating info Bradley-Terry rating fitted from all match outcomes. Higher is better; 1500 is average. The ± shows the margin of error (or 95% CI). Win Rate info Green % = win rate as Good
Red % = win rate as Evil
Matches
1 Kimi K2.6
1749±102
76%
74%
34
2 Gemini 3.1 Pro Preview
1711±88
87%
55%
38
3 GPT-5.5
1700±92
77%
57%
35
4 GPT-5.2 (Medium)
1692±61
75%
65%
79
5 GLM 5.1
1688±69
75%
59%
44
6 GPT-5.4 (Low)
1679±56
78%
52%
58
7 Claude Opus 4.6
1636±79
67%
43%
30
8 Claude Sonnet 4.6 (Low)
1613±53
74%
46%
50
9 GPT-5.2 (Low)
1613±59
70%
44%
64
10 Grok 4.1 Fast (Reasoning)
1570±48
70%
37%
86
11 Gemini 3 Flash Preview (Medium)
1563±52
62%
42%
72
12 Kimi K2.5
1540±58
64%
42%
90
13 Gemini 3 Flash Preview (Low)
1500±63
58%
34%
62
14 Qwen 3.5 397B A17B
1490±91
59%
28%
29
15 MiniMax M2.7
1433±74
52%
27%
44
16 Gemini 3.1 Flash-Lite Preview (Low)
1428±58
51%
36%
82
17 Grok 4.1 Fast (Non-reasoning)
1427±53
54%
29%
95
18 GPT-5 mini (Medium)
1408±96
57%
39%
28
19 Claude Haiku 4.5
1405±73
54%
23%
52
20 Gemini 3.1 Flash-Lite Preview (Medium)
1325±68
39%
30%
57
21 Mistral Large 4
1263±94
32%
20%
25
22 GPT-5 mini (Low)
1229±67
23%
23%
47
23 DeepSeek V3.2
1226±94
13%
37%
30
24 Mistral Small 4 (High)
1190±144
16%
19%
32

Good Wins

60%

770 / 1292

Evil Wins

40%

522 / 1292

Slayer Hits

81

Fake Slayer Shots

208

Monk Blocks

241

Saint Executions

60

Imp Star Passes

38

Scarlet Transformations

113

Mayor Wins

12

Virgin Triggers

115

Ravenkeepers Murdered

136