SocietyBench: leaderboard
SocietyBench (2026) asks LLMs to forecast how five real social events unfold after a cutoff date, from news and social-media timelines anonymized into counterfactual worlds so that memorized facts do not help. The overall score is the mean of a calibration axis and a temporal axis, each on a 0-100 scale where a trivial predictor scores 50; higher is better.
Metric: Overall score (0-100): mean of the calibration and temporal axes; five real social events (Wuhan Library, US-Iran, TikTok, SMCI, Trump Tariff) anonymized into counterfactual worlds, 25 prediction points each, averaged over the Chinese and English editions; temperature 0; trivial anchor 50. Source: arxiv.org. Saturation forecast: Around February 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 75 | #49 |
| 2 | Gemini 3.5 Flash | 70.6 | #101 |
| 3 | Claude Opus 4.8 | 66.3 | #56 |
| 4 | DeepSeek V4 Pro | 63.7 | #188 |
| 5 | Kimi K2.5 | 63.4 | #256 |
| 6 | Seed 2.0 Pro | 63.4 | #211 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=societybench · How It Works · Data refreshed daily, snapshot 2026-11-13.