Persian Hate Speech Reasoning - Content triage: leaderboard
Table 3; Persian multi-hop benchmark, one-shot CoT, temperature 0, top-p 1; Gemini reference annotations checked against human gold subset; task-specific reference agreement, not endorsement of content. Metric: Macro F1 (0–100). Higher is better. Source: https://arxiv.org/html/2610.20370v1
Metric: Macro F1 (0–100). Source: arxiv.org. Saturation forecast: Around January 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4.6 | 82.8 | #26 |
| 2 | GPT-5.2 | 79.7 | #109 |
| 3 | Claude Sonnet 4.6 | 79.2 | #89 |
| 4 | O3 | 77.8 | #123 |
| 5 | DeepSeek R1 | 75 | #246 |
| 6 | DeepSeek V3 | 68.5 | #314 |
| 7 | Qwen 3 30B A3B 2507 (Thinking) | 66.4 | #368 (Qwen 3 30B A3B 2507) |
| 8 | Qwen 3 30B A3B 2507 Instruct | 48 | #481 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #7 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=persian-hate-speech-reasoning-content-triage · How It Works · Data refreshed daily, snapshot 2026-10-16.