List your company

Response Harmfulness Classification on WildGuard (test)

79.48F1 (Total)GuardReasoner-VL 7B
14.396831.293448.1965.0866Jun 26, 2024Oct 1, 2024Jan 7, 2025Apr 15, 2025Jul 22, 2025Oct 28, 2025Feb 3, 2026

Evaluation results

MethodLinks
GuardReasoner-VL 7BModel Category=VLM Guard Model2026.0279.48———
Qwen3Guard Gen 8BModel Category=LLM Guard Model, Detection Mode=loose2026.0278.83———
GuardReasoner 8BModel Category=LLM Guard Model2026.0278.2———
Qwen3Guard Gen 8BModel Category=LLM Guard Model, Detection Mode=strict2026.0278.06———
PolyGuard-Qwen 7BModel Category=LLM Guard Model2026.0277.89———
GuardReasoner-Omni 4BModel Category=VLM Guard Model2026.0277.57———
GPT-42024.0677.373.681.3—
MD-Judge2024.0676.867.785—
MD-Judge 7BModel Category=LLM Guard Model2026.0276.8———
GuardReasoner-Omni 2BModel Category=VLM Guard Model2026.0276.4———
GuardReasoner-VL 3BModel Category=VLM Guard Model2026.0276.12———
WILDGUARD2024.0675.468.481.5—
WildGuard 7BModel Category=LLM Guard Model2026.0275.4———
LLaMA Guard 3 8BModel Category=LLM Guard Model2026.0270.8———
Llama-Guard22024.0666.547.978.2—
LLaMA Guard 4 12BModel Category=VLM Guard Model2026.0266.39———
BeaverDam2024.0663.451.274.3—
BeaverDam 7BModel Category=LLM Guard Model2026.0263.4———
LibrAI-LongFormer-harmbackbone=LongFormer2024.0663.261.764.2—
HarmB-Mistralbackbone=Mistral2024.0660.151.870.3—
HarmBench Mistral 7BModel Category=LLM Guard Model2026.0260.1———
Aegis-Guard-Ptype=Permissive2024.0656.44962.4—
Aegis Guard Permissive 7BModel Category=LLM Guard Model2026.0256.4———
Llama-Guard2024.0650.525.866.7—
Aegis-Guard-Dtype=Defensive2024.0649.140.457.6—
Aegis Guard Defensive 7BModel Category=LLM Guard Model2026.0249.1———
ShieldGemma 9BModel Category=LLM Guard Model2026.0247———
HarmB-Llamabackbone=Llama2024.0645.741.950.2—
HarmBench LLaMA 13BModel Category=LLM Guard Model2026.0245.7———
OAI Mod. API2024.0616.914.718.8—
Aegis Guard 2.0Model Size=8B2026.05———77.5
Aegis Guard DefensiveModel Size=7B2026.05———49.1
Aegis Guard PermissiveModel Size=7B2026.05———56.4
BeaverDamModel Size=7B2026.05———63.4
COLAGUARDModel Size=3B2026.05———77.23
COLAGUARDModel Size=8B2026.05———81.23
GPT-4oModel Size=Unknown2026.05———65.24
GPT-4o+CoTModel Size=Unknown2026.05———71.43
GuardReasonerModel Size=1B2026.05———74.81
GuardReasonerModel Size=3B2026.05———79.7
GuardReasonerModel Size=8B2026.05———78.2
HarmBench LLaMAModel Size=13B2026.05———45.7
HarmBench MistralModel Size=7B2026.05———60.1
LLaMA GuardModel Size=7B2026.05———50.5
LLaMA Guard 2Model Size=8B2026.05———66.5
LLaMA Guard 3Model Size=8B2026.05———70.8
MD-JudgeModel Size=7B2026.05———76.8
o1-previewModel Size=Unknown2026.05———50
QwQ-previewModel Size=32B2026.05———17.56
ShieldGemmaModel Size=2B2026.05———20.13
ShieldGemmaModel Size=9B2026.05———47
WildGuardModel Size=7B2026.05———75.4