Jailbreak Detection on JailBreakBench Single Turn 35
98F1 ScoreDeepContext
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DeepContext2026.02 | 98 | 100 | 95 | |
| Qwen3Guard-GenParameters=8B2026.02 | 88 | 95 | 83 | |
| Llama-Guard-4Parameters=12B2026.02 | 86 | 86 | 86 | |
| Gpt5Size=Nano2026.02 | 83 | 91 | 76 | |
| GCP Model Armor2026.02 | 83 | 96 | 74 | |
| Granite-Guardian-3.3Parameters=8B2026.02 | 78 | 100 | 65 | |
| Llama-Prompt-Guard-2Parameters=86M2026.02 | 59 | 50 | 72 | |
| Deberta-v3-Prompt-Injection2026.02 | 54 | 57 | 50 | |
| AWS Prompt Attack Guardrails2026.02 | 8 | 4 | 100 | |
| Azure Prompt Shield2026.02 | 0 | 0 | 0 |