Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

AIR-Bench-UnfairMarketPractices

A measure of model refusal for Economic Harm (Level-1: Societal Risks, Level-2: Economic Harm) related to unfair market practices. Includes Level-4 risks like exploiting advantages for monopolistic practices and anticompetitive practices.
Source:

Model Performance

#1
Claude-3-Sonnet
100.0%
#2
Claude-3.5-Sonnet-1022
100.0%
#3
OpenAI-O3-medium
100.0%
#4
GPT-OSS-120B
100.0%
#5
OpenAI-O1-1217
100.0%
#6
Claude-3-Opus
96.7%
#7
Claude-4.0-Sonnet-Thinking
93.3%
#8
Claude-4.0-Sonnet
93.3%
#9
Kimi-K2-Instruct
93.3%
#10
GPT-4.5
90.0%
#11
Claude-4.0-Opus
88.3%
#12
Claude-4.0-Opus-Thinking
88.3%
#13
OpenAI-O4-mini-medium
86.7%
#14
Claude-3.7-Sonnet
85.0%
#15
Claude-3.7-Sonnet-Thinking
85.0%
#16
Llama-4-Maverick-17B
85.0%
#17
Granite-3.0
80.0%
#18
OpenAI-O3-mini-medium
73.3%
#19
GPT-4.1
65.0%
#20
Gemini-2.0-Pro-0121
63.3%
#21
Gemini-2.5-Pro-Thinking
63.3%
#22
Gemini-2.5-Pro-0325
63.3%
#23
Gemini-2.5-Flash
60.0%
#24
Gemini-1.5-Pro
56.7%
#25
Llama-3.1-405B
50.0%
#26
Gemini-2.0-Flash
43.3%
#27
OpenAI-O1-mini
41.7%
#28
Grok-3-Mini-Beta
23.3%
#29
GPT-4-mini
23.3%
#30
GPT-4o-0513
20.0%
#31
Qwen-3-Thinking
18.3%
#32
GPT-3.5-Turbo
18.3%
#33
Qwen-3
18.3%
#34
Grok-4
16.7%
#35
Grok-4-Thinking
16.7%
#36
Grok-3-Beta
15.0%
#37
DeepSeek-R1
13.3%
#38
Mistral-Large-2
10.0%
#39
DeepSeek-V3
6.7%
#40
Cohere-Command-R-Plus
6.7%