ThaiSafetyBench

Thai language and Thai cultural-context safety benchmark reporting attack success rate across harmful-content categories.

24rows
overall_asr_pctprimary metric
2026-05-28sampled

Metadata

Metrics

Overall ASR (lower is better), Discrimination/Toxicity ASR (lower is better), Human-Chatbot Interaction Harms ASR (lower is better), Information Hazards ASR (lower is better), Malicious Uses ASR (lower is better), Misinformation Harms ASR (lower is better), Thai Socio-Cultural Harm ASR (lower is better), Thai Culture Related Attack ASR (lower is better), General Prompt Attack ASR (lower is better)

Latest Results

Rows are imported from the ThaiSafetyBench leaderboard results dataset and ranked by lowest overall ASR.

Rank Subject Overall ASR Model Match Provenance Sampled
1 GPT-5 4.43% overall ASR GPT-5
openai-gpt-5
Imported 2026-05-28
2 Claude 4.5 Sonnet 9.75% overall ASR Imported 2026-05-28
3 SeaLLMs-v3-7B 9.83% overall ASR Imported 2026-05-28
4 Qwen2.5-72B-Instruct 10.99% overall ASR Qwen2.5 72B Instruct
qwen-qwen-2.5-72b-instruct
Imported 2026-05-28
5 openthaigpt1.5-72b-instruct 12.34% overall ASR Imported 2026-05-28
6 Llama-SEA-LION-v3-70B 12.70% overall ASR Imported 2026-05-28
7 Qwen2.5-7B-Instruct 14.43% overall ASR Qwen2.5 7B Instruct
qwen-qwen-2.5-7b-instruct
Imported 2026-05-28
8 SeaLLMs-v3-1.5B 14.61% overall ASR Imported 2026-05-28
9 GPT-4o 16.04% overall ASR GPT-4o
openai-gpt-4o
Imported 2026-05-28
10 openthaigpt1.5-7b-instruct 16.09% overall ASR Imported 2026-05-28
11 typhoon2.1-gemma3-12b 16.64% overall ASR Imported 2026-05-28
12 Llama-3.3-70B-Instruct 16.87% overall ASR Llama 3.3 70B Instruct
meta-llama-llama-3.3-70b-instruct
Imported 2026-05-28
13 Llama-SEA-LION-v3-8B-IT 16.90% overall ASR Imported 2026-05-28
14 llama3.1-typhoon2-70b-instruct 18.05% overall ASR Imported 2026-05-28
15 gemma-3-12b-it 20.40% overall ASR Gemma 3 12B
google-gemma-3-12b-it
Imported 2026-05-28
16 typhoon2.1-gemma3-4b 22.97% overall ASR Imported 2026-05-28
17 Llama-3.1-70B-Instruct 24.49% overall ASR Llama 3.1 70B Instruct
meta-llama-llama-3.1-70b-instruct
Imported 2026-05-28
18 Llama-3.2-3B 26.08% overall ASR Imported 2026-05-28
19 gemma-3-4b-it 28.11% overall ASR Gemma 3 4B
google-gemma-3-4b-it
Imported 2026-05-28
20 Llama-3.1-8B-Instruct 28.24% overall ASR Llama 3.1 8B Instruct
meta-llama-llama-3.1-8b-instruct
Imported 2026-05-28
21 llama3.1-typhoon2-8b-instruct 32.44% overall ASR Imported 2026-05-28
22 llama3.2-typhoon2-3b-instruct 34.33% overall ASR Imported 2026-05-28
23 Llama-3.2-1B 37.66% overall ASR Imported 2026-05-28
24 llama3.2-typhoon2-1b-instruct 49.35% overall ASR Imported 2026-05-28