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	Kimi K2.6	1770±120	83% 78%	23
2	GPT-5.2 (Medium)	1720±62	78% 66%	76
3	Gemini 3.1 Pro Preview	1700±89	85% 56%	34
4	GLM 5.1	1698±72	74% 64%	39
5	GPT-5.4 (Low)	1695±56	78% 53%	55
6	Claude Opus 4.6	1647±78	68% 48%	25
7	GPT-5.2 (Low)	1638±59	72% 46%	61
8	Claude Sonnet 4.6 (Low)	1633±54	77% 47%	47
9	Grok 4.1 Fast (Reasoning)	1591±47	71% 38%	85
10	Gemini 3 Flash Preview (Medium)	1579±53	63% 41%	70
11	Kimi K2.5	1564±58	67% 44%	87
12	Gemini 3 Flash Preview (Low)	1520±63	61% 34%	59
13	Qwen 3.5 397B A17B	1502±94	59% 28%	29
14	MiniMax M2.7	1458±78	55% 29%	42
15	Gemini 3.1 Flash-Lite Preview (Low)	1443±58	53% 37%	77
16	Grok 4.1 Fast (Non-reasoning)	1440±52	54% 29%	92
17	GPT-5 mini (Medium)	1424±96	57% 39%	28
18	Claude Haiku 4.5	1418±73	53% 24%	51
19	Gemini 3.1 Flash-Lite Preview (Medium)	1342±65	39% 30%	57
20	Mistral Large 4	1286±95	35% 22%	23
21	GPT-5 mini (Low)	1249±66	24% 24%	45
22	DeepSeek V3.2	1243±93	13% 37%	30
23	Mistral Small 4 (High)	1210±138	17% 20%	30

Good Wins

60%

706 / 1182

Evil Wins

40%

476 / 1182

Slayer Hits

Fake Slayer Shots

198

Monk Blocks

226

Saint Executions

Imp Star Passes

Scarlet Transformations

108

Mayor Wins

Virgin Triggers

105

Ravenkeepers Murdered

125