FAR.AI AI Security Leaderboard: leaderboard
FAR.AI's red team attacks each frontier model's deployed safeguards with a fixed set of public jailbreak methods and their compositions, and the score is how many universal jailbreaks (attacks that work across a hazard category) it found. Unlike jailbreak success-rate boards such as Enkrypt or F5 CASI, it counts reusable breaks found by a budgeted search, so most current frontier models tie at zero.
Metric: Universal jailbreaks found (count; FAR.AI red team under the Minimal Standard for Safeguards v1.0, random-search and expert-guided attacks over chemical, biological, radiological & nuclear, explosives and cyber hazards; 0 = none found within a search budget above $14K). Source: leaderboard.far.ai. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol | 0 | #30 |
| 2 | GPT-6 Astra | 0 | #6 |
| 3 | Claude Opus 5.5 | 0 | #5 |
| 4 | Claude Fable 5 | 0 | #27 |
| 5 | Grok 4.7 | 0 | #44 |
| 6 | Claude Fable 5.1 | 0 | #15 |
| 7 | Grok 4.6 | 39 | #45 |
| 8 | Gemini 3.1 Pro (Preview) | 249 | #94 |
Interactive version: theaggregate.ai/benchmark?slug=far-ai-ai-security-leaderboard · How It Works · Data refreshed daily, snapshot 2026-11-13.