Clocktower Bench (Blood on the Clocktower) (Clocktower Bench)
We show this table for reference; we do not rank on it.
Frontier models play the social deduction game Blood on the Clocktower against each other, one model as the town and one as the Demon's team. The rating combines catching liars (detection) with lying well (deception).
Benchmark score on Clocktower Bench — November 6, 2026
We compile the Clocktower Bench rows from benchmark-owner or independent runs. The table contains 11 models. GPT-6 Astra has the highest published score at 1859, but the available coverage and evaluation setups do not establish a market leader. We do not use these results to rank models overall.
GPT-6 Astra
OpenAI
GPT-6.1 Sol
OpenAI
Claude Fable 5.1
Anthropic
11 modelsReasoningCurrentDisplay onlyUpdated November 6, 2026
Benchmark score table (11 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | GPT-6 AstraOpenAI | 1859 | Not reported | Closed |
| 2 | GPT-6.1 SolOpenAI | 1734 | Not reported | Closed |
| 3 | Claude Fable 5.1Anthropic | 1595 | Not reported | Closed |
| 4 | GPT-5.6 SolOpenAI | 1591 | Not reported | Closed |
| 5 | Claude Opus 5.5Anthropic | 1549 | Not reported | Closed |
| 6 | Claude Fable 5Anthropic | 1544 | Not reported | Closed |
| 7 | GPT-5.5OpenAI | 1482 | Not reported | Closed |
| 8 | Claude Opus 4.6Anthropic | 1432 | Not reported | Closed |
| 9 | GPT-5.4OpenAI | 1276 | Not reported | Closed |
| 10 | Claude Opus 4.7 (Adaptive)Anthropic | 1241 | Not reported | Closed |
| 11 | Claude Opus 4.8Anthropic | 1198 | Not reported | Closed |
How to read this leaderboard
Higher is better. A gap of about 200 points means the stronger model is expected to win roughly three games in four. Neighbours whose intervals overlap are not reliably ordered: GPT-5.6 Sol, Claude Opus 5.5 and Claude Fable 5 sit within a few points of one another.
Operator receipt: 11 sourced rows are currently displayable on this page; the highest published score among these 11 models is GPT-6 Astra at 1859, which does not establish a market leader.
Honest limit: Only OpenAI and Anthropic models are rated, so the scale is relative to this 11-model pool and shifts as models are added. Each model has 20 to 57 games per side, so intervals are wide (often more than 150 points). Claude models ran with adaptive thinking at high effort; the result does not enter overall or category rankings.
Among the reported Clocktower Bench rows, GPT-6 Astra is first at 1859. The third row is 264 score units behind. The broader top-10 range is 618 score units, so the table still separates the published systems.
11 models have been evaluated on Clocktower Bench. The benchmark falls in the Reasoning category. We keep these results separate from the weighted global scoring system, so they remain source-specific evidence. Clocktower Bench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About Clocktower Bench
Year
2026
Tasks
386 rated head-to-head games of Blood on the Clocktower (10 to 12 players)
Format
Relative rating from side-swapped game pairs
Difficulty
Social deduction, deception and multi-agent reasoning
Each game pits one model playing every good seat against another playing every evil seat, and each deal is replayed with the sides swapped. A single Bradley-Terry fit predicts the winner from both models' ratings, a good-side advantage and the setup (which Minions are in play; 10, 11 or 12 players). Ratings are in Elo points, centred so that 1500 is the average of the 11 rated models, with 90% intervals from a bootstrap over side-swapped pairs. The source also publishes separate detection (as good) and deception (as evil) ratings. Every model plays at its lab's high reasoning effort with an 8,000-token ceiling per turn. The rated set is the 2026 study: 386 games across 11 OpenAI and Anthropic models released in 2026.
Freshness and provenance
Version
Clocktower Bench 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does Clocktower Bench measure?
How well a model plays Blood on the Clocktower, a social deduction game where a hidden evil team lies to the town. Playing good tests detection: weighing claims and catching liars. Playing evil tests deception: keeping a cover story straight and coordinating with teammates. The overall rating combines both.
Which model leads Clocktower Bench?
In the 2026 study GPT-6 Astra leads at 1859, ahead of GPT-6.1 Sol at 1734. Claude Fable 5.1 is the top Anthropic model at 1595. OpenAI's lead comes mostly from detection: its town players keep updating on evidence across days, while Claude's towns tend to settle early.
Why are the ratings not percentages?
The score is a relative rating fitted from head-to-head games, with 1500 set to the average of the 11 rated models. It only compares models inside this pool, so a new model's arrival can shift every number.
Compare top models on Clocktower Bench
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 6,100+ subscribers.
One email each week. Unsubscribe anytime.