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	1755±75	89% 50%	36
2	Kimi K2.6	1728±84	73% 64%	44
3	GPT-5.2 (Medium)	1700±57	75% 61%	89
4	GPT-5.5	1693±81	73% 52%	44
5	Gemini 3.1 Pro Preview	1687±72	77% 50%	48
6	GLM 5.1	1678±71	73% 58%	45
7	GPT-5.4 (Low)	1673±54	78% 51%	59
8	Claude Opus 4.6	1636±77	68% 42%	31
9	Claude Sonnet 4.6 (Low)	1626±58	75% 46%	52
10	GPT-5.2 (Low)	1616±59	71% 45%	65
11	Grok 4.1 Fast (Reasoning)	1567±48	69% 37%	87
12	Gemini 3 Flash Preview (Medium)	1561±52	62% 41%	73
13	Kimi K2.5	1534±58	64% 42%	91
14	Gemini 3 Flash Preview (Low)	1499±63	58% 34%	62
15	Qwen 3.5 397B A17B	1487±92	59% 28%	29
16	MiniMax M2.7	1431±77	51% 27%	45
17	Grok 4.1 Fast (Non-reasoning)	1430±54	54% 29%	96
18	Gemini 3.1 Flash-Lite Preview (Low)	1428±58	51% 36%	82
19	GPT-5 mini (Medium)	1420±99	59% 38%	29
20	Claude Haiku 4.5	1405±72	54% 23%	52
21	Gemini 3.1 Flash-Lite Preview (Medium)	1327±67	39% 30%	57
22	Mistral Large 4	1277±91	31% 23%	26
23	GPT-5 mini (Low)	1229±67	23% 23%	48
24	DeepSeek V3.2	1227±92	13% 37%	30
25	Mistral Small 4 (High)	1205±127	18% 18%	33

Good Wins

60%

826 / 1376

Evil Wins

40%

550 / 1376

Slayer Hits

Fake Slayer Shots

216

Monk Blocks

264

Saint Executions

Imp Star Passes

Scarlet Transformations

120

Mayor Wins

Virgin Triggers

131

Ravenkeepers Murdered

145