Share your thoughts, 1 month free Claude Pro on usSee more

Safety Classification on ToxicChat

0.81F1 Score

Qwen3Guard

Updated 29d ago

Evaluation Results

Method	Links
Qwen3Guard 2025.12		0.81
Label and intent reward 2026.06		0.743
Nemotron Safety 4B 2026.06		0.733
LE-DPO 2026.06		0.733
AprielGuard 2025.12		0.73
Gemma-3-12B 2026.06		0.727
AprielGuard 2025.12		0.72
IBM Granite Guardian 3.3 2025.12		0.71
IF-DPO 2026.06		0.708
ShieldGemma 27B 2026.06		0.703
Human-intent on AIMS 2026.06		0.702
gpt-oss-safeguard 2025.12		0.7
Label reward 2026.06		0.685
GuardReasoner 8B 2026.06		0.681
IBM Granite Guardian 3.2 2025.12		0.68
GPT-5.4 2026.06		0.676
ShieldGemma 2025.12		0.67
Claude Sonnet 4.6 2026.06		0.667
Llama-3.1-8B 2026.06		0.664
WildGuard 7B 2026.06		0.652
Gemma-3-12B 2026.06		0.644
GPT-OSS-Safeguard 120B 2026.06		0.643
GPT-OSS-120B 2026.06		0.641
Qwen3Guard 2025.12		0.63
IBM Granite Guardian 3.3 2025.12		0.63
Llama-3.1-8B 2026.06		0.516
IBM Granite Guardian 3.1 2025.12		0.51
Llama Guard 3 2025.12		0.48
IBM Granite Guardian 3.2 2025.12		0.45
LlamaGuard 4 2026.06		0.441
Llama Guard 4 2025.12		0.43
Llama Guard 2 2025.12		0.42