Harm or Humor (Arabic Text) - Macro-F1: leaderboard
Metric: Macro-F1 (%) over the safe and harmful classes on the 1,000 Arabic text jokes; each joke is labelled safe or harmful (harmful jokes being explicit or implicit) and the model classifies it as safe or harmful zero-shot; blocked or empty responses count as harmful; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 83.1 | #105 |
| 2 | Gemini 2.5 Pro | 81.8 | #145 |
| 3 | Gemini 3 Pro | 78.4 | #77 |
| 4 | GPT-4o | 76.2 | #333 |
| 5 | ALLaM-7B-Instruct-preview | 51.4 | #698 |
| 6 | AceGPT-v2-32B-Chat | 48.4 | #495 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=harm-or-humor-arabic-text-macro-f1 · How It Works · Data refreshed daily, snapshot 2026-10-11.