Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

AIR-Bench-Erotic

A measure of model refusal for Sexual Content (Level-1: Content Safety Risks, Level-2: Sexual Content) related to erotic content. Includes Level-4 risks like erotic chats and fetishes.
Source:

Model Performance

#1
Claude 3 Opus
100.0%
#2
Claude 3.5 Haiku
100.0%
#3
Claude 4.0 Sonnet
100.0%
#4
Claude Sonnet 4.5
100.0%
#5
Claude 4.0 Sonnet (Thinking)
100.0%
#6
Claude Sonnet 4.5 (Thinking)
100.0%
#7
Qwen 3
100.0%
#8
Gemini 2.5 Flash (Thinking)
100.0%
#9
Gemini 2.5 Flash
100.0%
#10
Qwen 3 (Thinking)
100.0%
#11
GPT-4.5
97.5%
#12
Claude Haiku 4.5
95.2%
#13
Claude Haiku 4.5 (Thinking)
95.2%
#14
Claude 3.5 Sonnet
95.0%
#15
Granite 3.0
90.5%
#16
OpenAI o4 Mini (Medium Effort)
90.0%
#17
OpenAI o4 Mini (High Effort)
90.0%
#18
Claude 3 Sonnet
86.9%
#19
Claude 4.0 Opus
84.0%
#20
Claude 4.0 Opus (Thinking)
84.0%
#21
GPT OSS 120B
79.8%
#22
Llama 4 Maverick 17B
75.4%
#23
Claude 3.7 Sonnet (Thinking)
75.0%
#24
Claude 3.7 Sonnet
75.0%
#25
Gemini 3.1 Pro Preview
66.0%
#26
GPT-5 (Thinking)
64.3%
#27
GPT-5
64.3%
#28
OpenAI o1 Mini
63.8%
#29
Gemini 2.0 Pro
59.5%
#30
GPT-3.5 Turbo
52.4%
#31
Gemini 1.5 Pro
52.2%
#32
Grok 3
48.8%
#33
Grok 3 (Thinking)
48.8%
#34
OpenAI o1
48.6%
#35
OpenAI o3 Mini (Medium Effort)
46.4%
#36
Gemini 2.5 Pro
46.4%
#37
Gemini 2.5 Pro (Thinking)
46.4%
#38
OpenAI o3 Mini (High Effort)
46.4%
#39
OpenAI o3 (Medium Effort)
45.2%
#40
OpenAI o3 (High Effort)
45.2%
#41
Grok 4
42.0%
#42
Grok 4 (Thinking)
42.0%
#43
Grok 3 Mini
40.5%
#44
Kimi K2
36.7%
#45
Gemini 2.0 Flash Thinking (Jan 2025)
35.0%
#46
Gemini 2.0 Flash (Thinking)
35.0%
#47
Gemini 2.0 Flash
35.0%
#48
GPT-4o
28.3%
#49
Llama 3.1 405B
28.3%
#50
Llama 3.3 70B
28.3%
#51
Mistral Large 2
25.0%
#52
DeepSeek V3
21.7%
#53
DeepSeek R1
21.7%
#54
GPT-4o Mini
20.0%
#55
GPT-4.1
18.3%
#56
Cohere Command R+
16.7%
#57
Cohere Command A
15.0%