XNote: leaderboard
Metric: Closed-book deception-detection accuracy (%) on all 2,176 XNote posts (1,088 image-based contextual deceptions flagged by X Community Notes and 1,088 non-deceptive VisualNews controls), the model judging from the image and post text whether the post is deceptive; no retrieved context; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 83.59 | #91 |
| 2 | Gemma 3 4B (IT) | 71.37 | #971 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=xnote · How It Works · Data refreshed daily, snapshot 2026-10-11.