Safety Risk Detection on internal Agentic AI workflow benchmark benchmark leaderboard · SOTA2 Research

Safety Risk Detection on internal Agentic AI workflow benchmark

100PrecisionShieldGemma-9B

Evaluation Results

Method	Links
ShieldGemma-9B2025.12		100	3	5	0
Qwen3Guard-8B-loose2025.12		100	5	10	0
IBM-Granite-Guardian-3.3-8B2025.12		98	28	43	1
AprielGuard-8B2025.12		98	77	86	2
AprielGuard-8B2025.12		98	67	80	2
IBM-Granite-Guardian-3.3-8B2025.12		97	23	37	1
gpt-oss-safeguard-20B2025.12		97	46	62	2
Qwen3Guard-8B-strict2025.12		96	25	40	2
IBM-Granite-Guardian-3.1-2B2025.12		95	15	26	1
Llama-guard-3-8B2025.12		86	23	36	5
Llama-guard-4-12B2025.12		79	21	33	32
IBM-Granite-Guardian-3.2-3B2025.12		65	84	73	64
IBM-Granite-Guardian-3.2-5B2025.12		61	86	71	83
Llama-Prompt-Guard-2-86M2025.12		60	75	67	71
Llama-guard-3-8B2025.12		0.94	0.68	0.79	0.05
Llama-guard-4-12B2025.12		0.92	0.67	0.77	0.06
Llama-guard-2-8B2025.12		0.91	0.71	0.8	0.07
Qwen3Guard-8B-loose2025.12		0.91	0.87	0.89	0.09
ShieldGemma-9B2025.12		0.88	0.52	0.65	0.07
IBM-Granite-Guardian-3.2-5B2025.12		0.87	0.86	0.86	0.13
AprielGuard-8B2025.12		0.87	0.89	0.88	0.11
AprielGuard-8B2025.12		0.87	0.87	0.87	0.12
IBM-Granite-Guardian-3.3-8B2025.12		0.85	0.9	0.87	0.15
gpt-oss-safeguard-20B2025.12		0.85	0.86	0.85	0.14
IBM-Granite-Guardian-3.3-8B2025.12		0.84	0.88	0.86	0.16
IBM-Granite-Guardian-3.1-2B2025.12		0.83	0.86	0.84	0.17
Qwen3Guard-8B-strict2025.12		0.83	0.94	0.88	0.18
IBM-Granite-Guardian-3.2-3B2025.12		0.8	0.86	0.83	0.21
Llama-Prompt-Guard-2-86M2025.12		0.48	0.02	0.04	0.02