List your company

Full-response Safety Guardrail Classification on BeaverTails (test)

84.3F1 ScoreQwen3Guard-8B
67.03671.5187680.482Jun 1, 2026

Evaluation results

MethodLinks
Qwen3Guard-8B2026.0684.3
WildGuard-7B2026.0683.8
GPT-5.5protocol=zero-shot2026.0681.8
SentGuardbackbone=Qwen3-4B-Instruct-25072026.0681.2
Gemini-3.5-Flashprotocol=zero-shot2026.0670.2
LlamaGuard4-12B2026.0669.8
LlamaGuard3-8B2026.0667.7