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	GPT-5.2 (Medium)	1736±64	79% 67%	75
2	GLM 5.1	1706±78	74% 66%	38
3	GPT-5.4 (Low)	1704±59	80% 52%	54
4	Gemini 3.1 Pro Preview	1695±84	85% 55%	33
5	Claude Opus 4.6	1667±81	71% 50%	24
6	GPT-5.2 (Low)	1653±60	73% 47%	60
7	Claude Sonnet 4.6 (Low)	1641±55	78% 46%	46
8	Grok 4.1 Fast (Reasoning)	1601±46	71% 37%	84
9	Gemini 3 Flash Preview (Medium)	1592±54	64% 42%	69
10	Kimi K2.5	1573±60	66% 44%	86
11	Gemini 3 Flash Preview (Low)	1532±64	62% 34%	58
12	Qwen 3.5 397B A17B	1510±94	59% 28%	29
13	MiniMax M2.7	1464±80	55% 30%	40
14	Gemini 3.1 Flash-Lite Preview (Low)	1456±61	54% 38%	75
15	Grok 4.1 Fast (Non-reasoning)	1450±54	55% 30%	89
16	GPT-5 mini (Medium)	1435±97	57% 39%	28
17	Claude Haiku 4.5	1430±72	54% 24%	50
18	Gemini 3.1 Flash-Lite Preview (Medium)	1355±65	40% 31%	55
19	Mistral Large 4	1296±97	35% 22%	23
20	GPT-5 mini (Low)	1261±69	25% 25%	44
21	DeepSeek V3.2	1254±90	13% 37%	30
22	Mistral Small 4 (High)	1223±143	17% 21%	29

Good Wins

60%

682 / 1136

Evil Wins

40%

454 / 1136

Slayer Hits

Fake Slayer Shots

193

Monk Blocks

219

Saint Executions

Imp Star Passes

Scarlet Transformations

108

Mayor Wins

Virgin Triggers

103

Ravenkeepers Murdered

124