ROK-FORTRESS: leaderboard
Safety evaluations for large language models (LLMs) increasingly target high-stakes National Security and Public Safety (NSPS) risks, yet multilingual safety is typically assessed through translation-only benchmarks that preserve the underlying scenario, and empirical evidence of how language and geopolitical context interact remains limited to a narrow set of language pairs. We introduce R
Metric: Δ_ling (pp) (self-reported). Source: benchmarklist.com. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Mistral Large 3 | 28.4 | #401 |
| 2 | Llama 4 Maverick | 17.4 | #464 |
| 3 | DeepSeek V3 | 16.3 | #321 |
| 4 | Claude Sonnet 4.5 | 10.7 | #142 |
| 5 | Claude Opus 4.5 | 8.3 | #83 |
| 6 | Kimi K2 | 7.3 | #243 |
| 7 | GPT-5.2 | 4.3 | #110 |
| 8 | O4 Mini | 3.9 | #180 |
| 9 | Gemini 3 Pro | -0.9 | #80 |
| 10 | Qwen 3 235B A22B | -3 | #313 |
Interactive version: theaggregate.ai/benchmark?slug=rok-fortress · How It Works · Data refreshed daily, snapshot 2026-10-20.