Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
external benchmark
Clocktower Bench (Blood on the Clocktower). Every model the catalog carries a published Clocktower Bench value for, ranked by that value.
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). 12 models in the ModelChoice catalog have a published value.
Each point is a model with a published Clocktower Bench value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
12 models with a published Clocktower Bench value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Relative rating from side-swapped game pairs |
|---|---|---|---|
| 1 | OpenAI | 1859 | |
| 2 | OpenAI | 1734 | |
| 3 | Anthropic | 1595 |
What the benchmark measures and how its value is published, from the source’s own record.
Clocktower Bench (Clocktower Bench (Blood on the Clocktower)) is an external benchmark published by Clocktower Bench leaderboard. 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). Its evaluation set is described by the source as 386 rated head-to-head games of Blood on the Clocktower (10 to 12 players). The source rates its difficulty as social deduction, deception and multi-agent reasoning.
Each model's result is published by Clocktower Bench leaderboard as relative rating from side-swapped game pairs. ModelChoice prints that value to 0 decimal places, exactly as published, and never converts or rescales it. This page ranks the 12 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
| 4 | OpenAI | 1591 |
| 5 | Anthropic | 1549 |
| 6 | Anthropic | 1544 |
| 7 | OpenAI | 1482 |
| 8 | Anthropic | 1432 |
| 9 | OpenAI | 1276 |
| 10 | Anthropic | 1241 |
| 10 | Anthropic | 1241 |
| 12 | Anthropic | 1198 |
Evidence key: ObservedLast good
Rows are ordered by the value Clocktower Bench leaderboard published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.