SocialGrid (10x10 Rooms): leaderboard
Metric: Crewmate Elo rating (open scale, base 1500, K=64, 10 iterations) from a 30-matchup SocialGrid league in which one model plays all five crewmates and another both impostors in an Among Us-style gridworld (10x10 cells per room, 2x2 rooms, planning oracle on full assistance, no discussion phase); higher is better. Source: arxiv.org. Saturation forecast: Around November 2026. 6 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-OSS-120B | 1441 |
| 2 | DeepSeek R1 Distill Llama 70B | 1246 |
| 3 | Llama 3.1 70B Instruct | 1238 |
| 4 | Gemma 3 27B (IT) | 1168 |
Interactive version: theaggregate.ai/benchmark?slug=socialgrid-10x10-rooms · How It Works · Data refreshed daily, snapshot 2026-10-07.