Safety Classification on AegisSafetyTest V2
87F1 ScoreQwen3Guard
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3GuardSize=8B, Mode=strict, Reasoning=false2025.12 | 87 | — | |
| IBM Granite Guardian 3.3Size=8B, Reasoning=false2025.12 | 84 | — | |
| Qwen3GuardSize=8B, Mode=loose, Reasoning=false2025.12 | 84 | — | |
| AprielGuardSize=8B, Reasoning=false2025.12 | 84 | — | |
| IBM Granite Guardian 3.3Size=8B, Reasoning=true2025.12 | 83 | — | |
| IBM Granite Guardian 3.2Size=5B, Reasoning=false2025.12 | 82 | — | |
| AprielGuardSize=8B, Reasoning=true2025.12 | 82 | — | |
| gpt-oss-safeguardSize=20B, Reasoning=true2025.12 | 81 | — | |
| IBM Granite Guardian 3.1Size=2B, Reasoning=false2025.12 | 80 | — | |
| IBM Granite Guardian 3.2Size=3B, Reasoning=false2025.12 | 78 | — | |
| ShieldGemmaSize=9B, Reasoning=false2025.12 | 74 | — | |
| Llama Guard 2Reasoning=false2025.12 | 74 | — | |
| Llama Guard 3Reasoning=false2025.12 | 74 | — | |
| Llama Guard 4Reasoning=false2025.12 | 69 | — | |
| GLiGuard-LLMGuardrails-300M2026.05 | — | 81.61 | |
| Gliner-Guard-Omni2026.05 | — | 72.25 | |
| Nemotron Safety Guard v32026.05 | — | 84.33 | |
| Opir-edge2026.05 | — | 87.88 | |
| Opir-edge-multilang2026.05 | — | 93.21 | |
| Opir-multitask-large2026.05 | — | 93.08 | |
| Opir-multitask-multilang2026.05 | — | 86.71 | |
| PolyGuard-Qwen2026.05 | — | 83.79 | |
| PolyGuard-Qwen-Smol2026.05 | — | 82.78 | |
| Qwen3Guard-Gen-8B2026.05 | — | 82.49 | |
| WildGuard (vLLM)2026.05 | — | 75.31 |