Dwarkesh Patel releases the Blood on the Clocktower AI benchmark
Understand this story
Dwarkesh Patel built a multi-agent social deduction benchmark using Blood on the Clocktower and published model rankings. He said he spent about $20,000 in API credits to build the test framework, running a total of 386 matches; in each match, the good team consisted entirely of copies of one model, and the evil team consisted entirely of copies of another model. The results show that GPT-6 Astra, released in September 2026, is currently the strongest model, about 430 Elo higher than Claude Opus 4.6, released in February, with a win rate of about 92%. Patel released the code, invited others to verify it themselves, and said he would update the rankings at clocktowerbench.com as new models are released.
AI overview based on source reports · Updated 8 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Dwarkesh PatelThe benchmark where AIs have to lie to win
The author built a Blood on the Clocktower benchmark, spending about $20,000 in API credits on 386 games, to test AI social deception, multi-agent coordination and long-horizon reasoning. GPT-6 Astra ranked best, roughly 430 Elo points above Claude Opus 4.6, with OpenAI models notably stronger on the good side because they weigh evidence more skeptically. The code and a playable website are available at clocktowerbench.com.
Activity for this story
- Current comparable activity
- 8
- Peak comparable activity
- 10Nov 6 06:00
- Change in the past 24 hours
- –
Trends compare the same fully observed participants, which may cover fewer participants than current activity. Move the pointer or tap the chart to inspect hourly activity; use the left and right arrow keys with a keyboard.