Input Moderation on HarmBench (test)
100F1 ScoreQwen3Guard-4B-Gen-strict
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen3Guard-4B-Gen-strictStreaming=no2026.04 | 100 | |
| Qwen3Guard-8B-Gen-strictStreaming=no2026.04 | 100 | |
| Qwen3Guard-Stream-4B-strictStreaming=yes2026.04 | 100 | |
| Llama3-StreamGuard-8BStreaming=yes2026.04 | 99.7 | |
| Qwen3Guard-Stream-8B-strictStreaming=yes2026.04 | 99.4 | |
| Qwen3Guard-4B-Gen-looseStreaming=no2026.04 | 99.2 | |
| Llama3-StreamGuard-3BStreaming=yes2026.04 | 99.2 | |
| LlamaGuard3-8BStreaming=no2026.04 | 99 | |
| WildGuard-7BStreaming=no2026.04 | 98.9 | |
| Qwen3Guard-Stream-4B-looseStreaming=yes2026.04 | 98.9 | |
| PolyGuard-Qwen-7BStreaming=no2026.04 | 98.7 | |
| Qwen3Guard-0.6B-Gen-strictStreaming=no2026.04 | 98.7 | |
| Qwen3Guard-Stream-8B-looseStreaming=yes2026.04 | 98.7 | |
| Qwen3Guard-8B-Gen-looseStreaming=no2026.04 | 98.5 | |
| LlamaGuard4-12BStreaming=no2026.04 | 97.2 | |
| Qwen3Guard-Stream-0.6B-strictStreaming=yes2026.04 | 97.2 | |
| Qwen3Guard-Stream-0.6B-looseStreaming=yes2026.04 | 96.8 | |
| Qwen3Guard-0.6B-Gen-looseStreaming=no2026.04 | 96.1 | |
| Llama3-StreamGuard-1BStreaming=yes2026.04 | 94.8 | |
| NemoGuard-8BStreaming=no2026.04 | 75.2 | |
| ShieldGemma-9BStreaming=no2026.04 | 60.6 | |
| ShieldGemma-27BStreaming=no2026.04 | 57.3 |