Bench on the Clocktower
Ten models play the social deduction game Blood on the Clocktower together, one seat each, over 300 ten-player games of the Trouble Brewing script. Each model plays 210 games on the good team and 90 on the evil team. The score is the share of games its team won.
Every model plays in every game, so the good-team and evil-team rankings match by construction. Model names were hidden in half the games. The authors find a slight bias toward models from the same provider, which is not statistically significant. The post gives results only as charts, so we transcribed the win rates from its chart labels. The 300 game logs are public.
- Made by
- Tom Taylor and coauthors
- Released
- September 2026
- Kind
- Multi-agent games
- Level
- Everyday
- Scored
- Ranked by matches
- Answers
- Agent task
- Tasks
- All public
- Strongest model tested
- Claude Opus 5 #10 on the capability index
- Last model added
- July 2026
- Scores refreshed
- November 12th, 2026
GPT-5.6 Sol leads with 58% among models released by July 2026
Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best.
The newest model with a score here came out in July 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- Claude Fable 5.2
-
Gemini 4 Carbon - GPT-6 Astra
- Claude Opus 5.5
- and 12 more
- 1 GPT-5.6 Sol 58%
- 2 Claude Opus 5 55.3%
- 3 Claude Fable 5 52.7%
- 4 Gemini 3.1 Pro 52%
- 5 Grok 4.5 50%
- 6 Kimi K3 49.3%
- 7 GPT-5.6 Terra 48.7%
- 8
DeepSeek-V4-Pro 48% - 9 Gemini 3.6 Flash 46.7%
- 10
Qwen 3.6 35B-A3B 42%
Source: Official Bench on the Clocktower leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-5.6 Sol | 58% | Jul 2026 | Official leaderboard | |
| 2 | Claude Opus 5 | 55.3% | Jul 2026 | Official leaderboard | |
| 3 | Claude Fable 5 | 52.7% | Jun 2026 | Official leaderboard | |
| 4 | Gemini 3.1 Pro | 52% | Feb 2026 | Official leaderboard | |
| 5 | Grok 4.5 | 50% | Jul 2026 | Official leaderboard | |
| 6 | Kimi K3 | 49.3% | Jul 2026 | Official leaderboard | |
| 7 | GPT-5.6 Terra | 48.7% | Jul 2026 | Official leaderboard | |
| 8 | DeepSeek-V4-Pro | 48% | Apr 2026 | Official leaderboard | |
| 9 | Gemini 3.6 Flash | 46.7% | Jul 2026 | Official leaderboard | |
| 10 | Qwen 3.6 35B-A3B | 42% | Apr 2026 | Official leaderboard |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Bench on the Clocktower leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Bench on the Clocktower leaderboard
No lab model card reports it yet
We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.
More in Games and long tasks
- ARC-AGI-3 In use Interactive game environments where agents must explore, infer goals and rules, and act efficiently without instructions. GPT-6 Astra 99.95%
- Orak Fading LLM agents play 12 popular video games across most genres through an MCP interface. Each game scores in-game progress or outcomes, and the board ranks models by their average rank across the games. Gemini 2.5 Pro 3.90
- MineDojo Fading Agents complete open-ended tasks in Minecraft, such as mining blocks and fighting mobs, from a large simulation suite. Programmatic checks score many tasks.
Seed2.0 Pro 49% - TextQuests Only in system cards Agents play 25 classic interactive fiction games. The score is harmful actions per 100 steps, where lower is better. The board also reports game progress.
Llama 4 Scout 7.00 - GameWorld Only in system cards Multimodal game agents play 34 browser games across 170 tasks. Outcomes are checked from the game state. GPT-6.1 Terra 76.8%
- Swarm Bench Only in system cards Models orchestrate agent swarms on complex tasks that benefit from coordinated decomposition and parallel execution. Kimi K3 76.3%