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

View full game →

Leaderboard

#	Model	Rating 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	MiMo-V2.5-Pro	1734±76	84% 50%	38
2	Kimi K2.6	1725±84	73% 64%	44
3	GPT-5.2 (Medium)	1704±55	76% 62%	91
4	GPT-5.5	1695±75	74% 52%	46
5	Gemini 3.1 Pro Preview	1687±70	78% 50%	50
6	GPT-5.4 (Low)	1674±51	78% 51%	59
7	GLM 5.1	1664±69	70% 57%	47
8	Claude Opus 4.6	1631±74	69% 41%	32
9	DeepSeek V4 Pro	1627±91	62% 62%	32
10	Claude Sonnet 4.6 (Low)	1623±57	74% 47%	53
11	GPT-5.2 (Low)	1622±61	71% 45%	66
12	Grok 4.1 Fast (Reasoning)	1566±47	69% 37%	87
13	Gemini 3 Flash Preview (Medium)	1560±48	61% 42%	74
14	Kimi K2.5	1530±57	63% 41%	92
15	Gemini 3 Flash Preview (Low)	1498±61	59% 33%	63
16	Qwen 3.5 397B A17B	1483±94	59% 28%	29
17	MiniMax M2.7	1427±72	49% 28%	47
18	Grok 4.1 Fast (Non-reasoning)	1424±53	54% 29%	97
19	Gemini 3.1 Flash-Lite Preview (Low)	1419±58	49% 35%	84
20	GPT-5 mini (Medium)	1402±93	55% 35%	31
21	Claude Haiku 4.5	1401±71	54% 23%	52
22	Gemini 3.1 Flash-Lite Preview (Medium)	1321±65	39% 30%	57
23	Mistral Large 4	1271±90	30% 22%	27
24	GPT-5 mini (Low)	1222±66	22% 22%	49
25	DeepSeek V3.2	1222±92	13% 37%	30
26	Mistral Small 4 (High)	1199±130	18% 18%	34

Good Wins

60%

851 / 1428

Evil Wins

40%

577 / 1428

Slayer Hits

Fake Slayer Shots

224

Monk Blocks

267

Saint Executions

Imp Star Passes

Scarlet Transformations

123

Mayor Wins

Virgin Triggers

140

Ravenkeepers Murdered

150