ChineseSafe Benchmark: leaderboard
Chinese content safety evaluation: measures LLM accuracy in identifying unsafe vs safe content in Chinese across multiple safety categories.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 55 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 32B | 75.26 | #443 |
| 2 | Llama 4 Maverick | 75.02 | #464 |
| 3 | Qwen 3 4B | 74.95 | #829 |
| 4 | GPT-4o | 73.78 | #343 |
| 5 | Phi-3-small-8k-instruct | 72.73 | #967 |
| 6 | Phi-4 | 72.24 | #706 |
| 7 | Qwen 2.5 3B Instruct | 71.81 | #1162 |
| 8 | Gemma 1.1 7B (IT) | 71.7 | #1368 |
| 9 | GPT-4 Turbo | 71.67 | #404 |
| 10 | deepseek-llm-7B-chat | 71.63 | #1354 |
| 11 | Gemma 3 4B (IT) | 71.41 | #992 |
| 12 | Gemini 2.5 Flash (Preview 05-20) | 71.27 | #216 |
| 13 | GLM-4 9B Chat | 70.96 | #916 |
| 14 | Mistral 7B Instruct (v0.3) | 70.41 | #1274 |
| 15 | Qwen 1.5 7B Chat | 70.36 | #1324 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=chinesesafe-benchmark · How It Works · Data refreshed daily, snapshot 2026-10-20.