# F5 Labs CASI & ARS Leaderboard

Tests how hard it is to trick AI models into harmful behavior using attack prompts. A higher score means the model is harder to break.

Category: safety. Type: leaderboard. Status: active.
Metric: CASI: F5's Comprehensive AI Security Index (higher = more secure) (higher is better).
Source: F5 Labs — https://www.f5.com/labs/casi, results as of 2026-11, retrieved 2026-11-17, license: not stated.

| Rank | Provider | Model | CASI | Avg. Performance |
|---|---|---|---|---|
| 1 | Anthropic | Claude Sonnet 5 | 98.00 | 23.20% |
| 2 | Anthropic | Claude Haiku 5.5 | 94.61 | 43.40% |
| 3 | xAI | Grok 4.7 High | 93.31 | 46.40% |
| 4 | OpenAI | GPT-5.4 | 93.21 | 18.20% |
| 5 | Anthropic | Claude Opus 5 | 93.05 | 50.80% |
| 6 | Qwen | Qwen 3.8-Max | 92.31 | 40.20% |
| 7 | Anthropic | Claude Fable 5.1 Max Effort | 92.30 | 53.40% |
| 8 | Anthropic | Claude Opus 5.5 Reasoning | 92.00 | 57.60% |
| 9 | Anthropic | Claude Sonnet 5.5 | 91.78 | 56.00% |
| 10 | OpenAI | GPT-6 Luna | 89.15 | 38.10% |

Notes: Only top 10 shown in the page's initial table; latest month selector shows November 2026. ARS view not extracted.
