LLMsPark - Who Is Spy: leaderboard
Models describe their hidden words and identify the minority player in a social-deduction game. The primary uses a tailored higher-better game score, not standard Elo; its finite range and denominator are unspecified.
Metric: Source-defined Who Is Spy game score (higher is better). Source: llmsparks.github.io. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen-14B-Chat | 77 | #1169 |
| 2 | GPT-3.5 Turbo | 60 | #865 |
| 3 | GPT-4 | 59 | #438 |
Interactive version: theaggregate.ai/benchmark?slug=llmspark-who-is-spy · How It Works · Data refreshed daily, snapshot 2026-10-20.