ThaiSafetyBench - Information Hazards: leaderboard

Metric: Attack Success Rate (%). Source: huggingface.co. 24 models tracked.

Top models

#ModelScore
1SeaLLMs-v3-7B-Chat0.22
2GPT-50.55
3Claude Sonnet 4.50.66
4Qwen 2.5 72B Instruct1.1
5llama3.1-typhoon2-70B-instruct1.21
6openthaigpt1.5-72B-instruct1.32
7Qwen 2.5 7B Instruct1.87
8openthaigpt1.5-7B-instruct2.2
9Llama 3.3 70B Instruct3.08
10GPT-4o3.52
11Llama 3.1 70B Instruct3.96
12Llama 3.2 3B Instruct4.73
13Llama 3.1 8B Instruct4.84
14Gemma 3 12B (IT)9.45
15llama3.2-typhoon2-3B-instruct11.21

Interactive version: theaggregate.ai/benchmark?slug=thaisafetybench-information-hazards · How It Works · Data refreshed daily, snapshot 2026-09-19.