Horangi 4 - Korean Hate Speech: leaderboard
Metric: Accuracy (%). Source: horangi.ai. 105 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Fable 5 (High) | 78 |
| 2 | Gemma 3 27B (IT) | 74 |
| 3 | EXAONE-3.0-7.8B-Instruct | 71 |
| 4 | Grok 4.1 Fast (Non-reasoning) | 70 |
| 5 | DeepSeek V3.2 (High) | 70 |
| 6 | Mistral Medium 3.5 | 69 |
| 7 | Grok 4.20 | 69 |
| 8 | GLM-5-Turbo | 69 |
| 9 | Claude Opus 4.6 (High) | 69 |
| 10 | Qwen 3.5 397B A17B | 67 |
| 11 | Mistral Small 4 | 67 |
| 12 | Claude Sonnet 4.6 (High) | 67 |
| 13 | A.X 4.0 | 67 |
| 14 | Gemma 3 12B (IT) | 66 |
| 15 | Qwen 3.7 Max | 66 |
Interactive version: theaggregate.ai/benchmark?slug=horangi-4-korean-hate-speech · How It Works · Data refreshed daily, snapshot 2026-09-19.