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 Opus
100.0%
#2
Claude 3.5 Haiku
100.0%
#3
Claude 3.5 Sonnet
100.0%
#4
Claude Haiku 4.5 (Thinking)
100.0%
#5
Claude Haiku 4.5
100.0%
#6
OpenAI o4 Mini (High Effort)
100.0%
#7
OpenAI o4 Mini (Medium Effort)
100.0%
#8
Gemini 2.5 Flash (Thinking)
100.0%
#9
Gemini 2.5 Flash
100.0%
#10
OpenAI o1
100.0%
#11
Llama 4 Maverick 17B
100.0%
#12
OpenAI o3 Mini (High Effort)
100.0%
#13
OpenAI o3 Mini (Medium Effort)
100.0%
#14
Claude Sonnet 4.5 (Thinking)
99.4%
#15
Claude Sonnet 4.5
99.4%
#16
Claude 3 Sonnet
99.4%
#17
GPT OSS 120B
97.5%
#18
GPT-5
95.8%
#19
GPT-5 (Thinking)
95.8%
#20
Granite 3.0
95.0%
#21
Claude 4.0 Opus
94.2%
#22
Claude 4.0 Opus (Thinking)
94.2%
#23
Gemini 3.1 Pro Preview
93.3%
#24
Kimi K2
93.3%
#25
Gemini 2.5 Pro
88.0%
#26
Gemini 2.5 Pro (Thinking)
88.0%
#27
OpenAI o3 (High Effort)
87.5%
#28
OpenAI o3 (Medium Effort)
87.5%
#29
Claude 3.7 Sonnet
85.0%
#30
Claude 3.7 Sonnet (Thinking)
85.0%
#31
Qwen 3 (Thinking)
76.4%
#32
Qwen 3
76.4%
#33
Gemini 1.5 Pro
73.6%
#34
Claude 4.0 Sonnet
67.0%
#35
Claude 4.0 Sonnet (Thinking)
67.0%
#36
GPT-4.1
65.0%
#37
GPT-4.5
60.0%
#38
Llama 3.1 405B
50.0%
#39
Gemini 2.0 Pro
45.7%
#40
Gemini 2.0 Flash Thinking (Jan 2025)
43.3%
#41
Gemini 2.0 Flash (Thinking)
43.3%
#42
Gemini 2.0 Flash
43.3%
#43
Grok 3 Mini
30.0%
#44
GPT-3.5 Turbo
30.0%
#45
Grok 3
28.9%
#46
Grok 3 (Thinking)
28.9%
#47
OpenAI o1 Mini
26.7%
#48
GPT-4o Mini
23.3%
#49
GPT-4o
20.0%
#50
Llama 3.3 70B
16.7%
#51
DeepSeek R1
13.3%
#52
Mistral Large 2
10.0%
#53
Cohere Command A
6.7%
#54
DeepSeek V3
6.7%
#55
Cohere Command R+
6.7%
#56
Grok 4
4.8%
#57
Grok 4 (Thinking)
4.8%