Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

AIR-Bench-Monetized

A measure of model refusal for Sexual Content (Level-1: Content Safety Risks, Level-2: Sexual Content) related to monetized sexual content. Includes Level-4 risks like pornography and promotion of sexual services.
Source:

Model Performance

#1
Claude-3.5-Sonnet-1022
90.0%
#2
Claude-3-Sonnet
86.7%
#3
Claude-3-Opus
81.7%
#4
Llama-4-Maverick-17B
76.7%
#5
Gemini-2.0-Pro-0121
68.3%
#6
Granite-3.0
66.7%
#7
GPT-4.5
61.7%
#8
GPT-OSS-120B
60.0%
#9
Claude-4.0-Sonnet
56.7%
#10
Claude-4.0-Sonnet-Thinking
56.7%
#11
Gemini-2.0-Flash
55.0%
#12
Claude-4.0-Opus-Thinking
53.3%
#13
Claude-4.0-Opus
53.3%
#14
Gemini-1.5-Pro
53.3%
#15
Claude-3.7-Sonnet-Thinking
43.3%
#16
Claude-3.7-Sonnet
43.3%
#17
Gemini-2.5-Pro-0325
41.7%
#18
Gemini-2.5-Pro-Thinking
41.7%
#19
Kimi-K2-Instruct
38.3%
#20
Gemini-2.5-Flash
28.3%
#21
OpenAI-O3-medium
26.7%
#22
OpenAI-O4-mini-medium
26.7%
#23
Qwen-3
23.3%
#24
GPT-3.5-Turbo
23.3%
#25
Qwen-3-Thinking
23.3%
#26
DeepSeek-R1
21.7%
#27
Llama-3.1-405B
21.7%
#28
GPT-4o-0513
20.0%
#29
OpenAI-O1-mini
18.3%
#30
Grok-3-Beta
18.3%
#31
Grok-3-Mini-Beta
18.3%
#32
GPT-4.1
16.7%
#33
Grok-4-Thinking
16.7%
#34
Grok-4
16.7%
#35
Cohere-Command-R-Plus
15.0%
#36
GPT-4-mini
15.0%
#37
Mistral-Large-2
13.3%
#38
OpenAI-O1-1217
13.3%
#39
DeepSeek-V3
11.7%
#40
OpenAI-O3-mini-medium
11.7%