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.
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 | Cost info Based on USD cost per game. | Verbosity info Based on output tokens per action. Mostly consumed by reasoning (if enabled). |
|---|---|---|---|---|---|
| 1 | MiMo-V2.5-Pro | 1733±70 | 83% 51% | ||
| 2 | Kimi K2.6 | 1712±80 | 70% 63% | ||
| 3 | GPT-5.2 (Medium) | 1699±58 | 76% 60% | ||
| 4 | GPT-5.5 | 1695±75 | 75% 52% | ||
| 5 | Gemini 3.1 Pro Preview | 1689±66 | 79% 49% | ||
| 6 | DeepSeek V4 Pro | 1689±85 | 67% 70% | ||
| 7 | GPT-5.4 (Low) | 1681±54 | 79% 52% | ||
| 8 | GLM 5.1 | 1666±66 | 71% 57% | ||
| 9 | GPT-5.2 (Low) | 1630±59 | 72% 47% | ||
| 10 | Claude Opus 4.6 | 1617±71 | 66% 40% | ||
| 11 | Claude Sonnet 4.6 (Low) | 1613±54 | 71% 46% | ||
| 12 | Grok 4.1 Fast (Reasoning) | 1562±46 | 67% 37% | ||
| 13 | Gemini 3 Flash Preview (Medium) | 1555±50 | 60% 42% | ||
| 14 | Grok 4.3 | 1538±63 | 57% 43% | ||
| 15 | Kimi K2.5 | 1529±55 | 63% 41% | ||
| 16 | Gemini 3 Flash Preview (Low) | 1494±62 | 58% 33% | ||
| 17 | Qwen 3.5 397B A17B | 1480±94 | 59% 28% | ||
| 18 | MiniMax M2.7 | 1422±70 | 49% 27% | ||
| 19 | Grok 4.1 Fast (Non-reasoning) | 1422±56 | 54% 28% | ||
| 20 | Gemini 3.1 Flash-Lite Preview (Low) | 1410±58 | 48% 34% | ||
| 21 | Claude Haiku 4.5 | 1397±71 | 54% 23% | ||
| 22 | GPT-5 mini (Medium) | 1386±94 | 52% 33% | ||
| 23 | Gemini 3.1 Flash-Lite Preview (Medium) | 1317±63 | 39% 30% | ||
| 24 | Mistral Large 4 | 1276±94 | 31% 21% | ||
| 25 | DeepSeek V3.2 | 1218±92 | 13% 37% | ||
| 26 | GPT-5 mini (Low) | 1213±65 | 22% 22% | ||
| 27 | Mistral Small 4 (High) | 1206±129 | 19% 17% |
MiMo-V2.5-Pro 
Kimi K2.6 
GPT-5.2 (Medium) 
Gemini 3.1 Pro Preview 
DeepSeek V4 Pro 
GLM 5.1 
Claude Opus 4.6 
Grok 4.1 Fast (Reasoning) 
Qwen 3.5 397B A17B 
MiniMax M2.7 
Mistral Large 4 Good Wins
59%
910 / 1532
Evil Wins
41%
622 / 1532
Slayer Hits
97
Fake Slayer Shots
241
Monk Blocks
286
Saint Executions
71
Imp Star Passes
41
Scarlet Transformations
132
Mayor Wins
14
Virgin Triggers
149
Ravenkeepers Murdered
165