Model Threat Profiles

Security ratings for 21 major LLMs. Known vulnerabilities, attack resistance, and threat profiles for every model you're deploying.

21 models
Anthropic

Claude Opus 4.6

86
Strong
Injection
86
Leakage
82
Instructions
90
Jailbreak
84
Output
86
View threat profile →
OpenAI

GPT-5.4

82
Strong
Injection
82
Leakage
78
Instructions
86
Jailbreak
80
Output
84
View threat profile →
Anthropic

Claude Sonnet 4.6

80
Strong
Injection
80
Leakage
78
Instructions
84
Jailbreak
78
Output
82
View threat profile →
Google

Gemini 3.1 Pro

80
Strong
Injection
80
Leakage
76
Instructions
84
Jailbreak
78
Output
80
View threat profile →
OpenAI

o3

78
Moderate
Injection
78
Leakage
72
Instructions
84
Jailbreak
76
Output
80
View threat profile →
Anthropic

Claude Opus 4

78
Moderate
Injection
78
Leakage
76
Instructions
82
Jailbreak
76
Output
80
View threat profile →
OpenAI

GPT-5.4 mini

76
Moderate
Injection
76
Leakage
72
Instructions
80
Jailbreak
74
Output
78
View threat profile →
Anthropic

Claude Sonnet 4

74
Moderate
Injection
74
Leakage
72
Instructions
78
Jailbreak
72
Output
76
View threat profile →
Google

Gemini 2.5 Pro

74
Moderate
Injection
74
Leakage
70
Instructions
78
Jailbreak
72
Output
76
View threat profile →
OpenAI

o3-mini

70
Moderate
Injection
70
Leakage
66
Instructions
76
Jailbreak
68
Output
72
View threat profile →
Anthropic

Claude Haiku 4.5

70
Moderate
Injection
70
Leakage
68
Instructions
74
Jailbreak
68
Output
72
View threat profile →
OpenAI

GPT-5.3 Instant

68
Moderate
Injection
68
Leakage
64
Instructions
72
Jailbreak
66
Output
70
View threat profile →
OpenAI

GPT-5.3 Codex

67
Moderate
Injection
66
Leakage
62
Instructions
74
Jailbreak
64
Output
70
View threat profile →
OpenAI

GPT-4o

64
Fair
Injection
64
Leakage
60
Instructions
70
Jailbreak
62
Output
66
View threat profile →
Google

Gemini 2.0 Flash

62
Fair
Injection
62
Leakage
58
Instructions
68
Jailbreak
60
Output
64
View threat profile →
xAI

Grok 4.20

62
Fair
Injection
62
Leakage
58
Instructions
68
Jailbreak
56
Output
64
View threat profile →
Meta

Llama 4

62
Fair
Injection
62
Leakage
56
Instructions
68
Jailbreak
58
Output
64
View threat profile →
Mistral

Mistral Large

62
Fair
Injection
62
Leakage
58
Instructions
68
Jailbreak
60
Output
64
View threat profile →
Alibaba

Qwen 3.5

59
Fair
Injection
58
Leakage
54
Instructions
66
Jailbreak
56
Output
62
View threat profile →
DeepSeek

DeepSeek V3.2

55
Fair
Injection
54
Leakage
50
Instructions
62
Jailbreak
52
Output
58
View threat profile →
Google

Gemini 2.0 Flash-Lite

49
Weak
Injection
48
Leakage
44
Instructions
54
Jailbreak
46
Output
52
View threat profile →

Want to test how your specific deployment holds up?

Scan Your Agent
Scan Agent