Frontier AI Risk Monitor - Fortress-Biological: leaderboard
Scale's FORTRESS adversarial prompts in the biological subdomain (31 prompts, 5 epochs) graded against per-prompt rubrics by DeepSeek V4 Flash; the platform reverses the risk score, so higher is safer.
Metric: Safety score (0-100, higher is safer). Source: airiskmonitor.net. 46 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 98.25 | #5 |
| 2 | Grok 4.7 | 98.25 | #44 |
| 3 | GPT-6 Astra | 98.16 | #6 |
| 4 | Claude Opus 4.8 | 97.6 | #56 |
| 5 | Claude Opus 4.6 (Thinking) | 96.11 | #102 (Claude Opus 4.6) |
| 6 | GPT-5.4 (High) | 95.74 | #123 (GPT-5.4) |
| 7 | GPT-5.5 | 95.6 | #49 |
| 8 | Gemini 3.8 Flash | 95.21 | #36 |
| 9 | Claude Opus 4.5 (Thinking) | 95.16 | #156 (Claude Opus 4.5) |
| 10 | GPT-5.6 Sol | 90.52 | #30 |
| 11 | GPT-5.1 (High) | 89.03 | #229 (GPT-5.1) |
| 12 | Grok 4.3 | 86.13 | #333 |
| 13 | MiniMax-M2 | 81.43 | #674 |
| 14 | MiniMax-M2.7 | 77.95 | #518 |
| 15 | Qwen 3.5 Plus (Thinking) | 77.49 | #232 (Qwen 3.5 Plus) |
Interactive version: theaggregate.ai/benchmark?slug=frontier-ai-risk-monitor-fortress-biological · How It Works · Data refreshed daily, snapshot 2026-11-13.