Evaluation profile
ThaiSafetyBench
1sub-evals
0.0667%total index weight
1components
Within-component eval weight: Misuse resistance 0.667%.
Model score (higher is better)Predicted score
About this eval
Resistance to Thai-language malicious prompts, including culturally contextualized attacks.
Included in the behavior ranking.
Sub-evals
| Measure | Component | Direction | Total index weight | Within-component weight |
|---|---|---|---|---|
| safety_scorethaisafetybench/thaisafetybench.csv:safety_scoreMeasures resistance to harmful Thai-language prompt attacks. | ordinary_harm_misuse_resistance:1.000do-not-answer/do-not-answer.csv | Higher is better | 0.0667% | Misuse resistance 0.667% |
safety_score
Measures resistance to harmful Thai-language prompt attacks.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | gpt-5 | 95.57 | official | |
| 2 | claude-sonnet-4.5 | 90.25 | official | |
| 3 | seallms-v3-7b-chat | 90.17 | official | |
| 4 | qwen2.5-72b-instruct | 89.01 | official | |
| 5 | openthaigpt1.5-72b-instruct | 87.66 | official | |
| 6 | llama-sea-lion-v3-70b-it | 87.3 | official | |
| 7 | qwen-2.5-7b-instruct | 85.57 | official | |
| 8 | seallms-v3-1.5b-chat | 85.39 | official | |
| 9 | gpt-4o | 83.96 | official | |
| 10 | openthaigpt1.5-7b-instruct | 83.91 | official | |
| 11 | llama-3.3-70b-instruct | 83.13 | official | |
| 12 | llama-sea-lion-v3-8b-it | 83.1 | official | |
| 13 | gemma-3-12b | 79.6 | official | |
| 14 | llama-3.1-70b-instruct | 75.51 | official | |
| 15 | llama-3.2-3b | 73.92 | official | |
| 16 | gemma-3-4b | 71.89 | official | |
| 17 | llama-3.1-8b-instruct | 71.76 | official | |
| 18 | llama-3.2-1b-instruct | 62.34 | official |