Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

AIR-Bench-SpecificTypesOfRights

A measure of model refusal for Fundamental Rights (Level-1: Legal and Rights-Related Risks, Level-2: Fundamental Rights) related to violating specific types of rights. Includes Level-4 risks like IP rights/trade secrets, likeness rights, reputational rights, honor, and name rights violations.
Source:

Model Performance

#1
Claude-3-Opus
100.0%
#2
Claude-3-Sonnet
100.0%
#3
Claude-3.5-Sonnet-1022
100.0%
#4
Claude-4.0-Sonnet
100.0%
#5
GPT-OSS-120B
100.0%
#6
Claude-4.0-Sonnet-Thinking
100.0%
#7
Claude-3.7-Sonnet
97.3%
#8
Claude-3.7-Sonnet-Thinking
97.3%
#9
OpenAI-O4-mini-medium
97.3%
#10
Granite-3.0
97.3%
#11
OpenAI-O3-medium
97.3%
#12
Claude-4.0-Opus
96.7%
#13
Claude-4.0-Opus-Thinking
96.7%
#14
OpenAI-O1-1217
96.0%
#15
OpenAI-O3-mini-medium
94.7%
#16
GPT-4.5
93.3%
#17
Kimi-K2-Instruct
92.0%
#18
Gemini-2.5-Pro-0325
89.3%
#19
Gemini-2.5-Flash
89.3%
#20
Gemini-2.5-Pro-Thinking
89.3%
#21
GPT-4.1
86.7%
#22
Llama-4-Maverick-17B
84.7%
#23
GPT-4-mini
84.7%
#24
Gemini-2.0-Pro-0121
83.3%
#25
Gemini-2.0-Flash
80.0%
#26
Gemini-1.5-Pro
76.0%
#27
GPT-4o-0513
73.3%
#28
Llama-3.1-405B
73.3%
#29
GPT-3.5-Turbo
69.3%
#30
Qwen-3
68.0%
#31
Qwen-3-Thinking
68.0%
#32
Grok-3-Mini-Beta
62.7%
#33
DeepSeek-R1
60.0%
#34
Grok-3-Beta
57.3%
#35
DeepSeek-V3
46.7%
#36
OpenAI-O1-mini
41.3%
#37
Grok-4-Thinking
40.0%
#38
Grok-4
40.0%
#39
Mistral-Large-2
32.7%
#40
Cohere-Command-R-Plus
20.0%