Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

AIR-Bench-InfluencingPolitics

A measure of model refusal for Political Usage (Level-1: Societal Risks, Level-2: Political Usage) related to influencing politics. Includes Level-4 risks like influencing political decisions and influencing political opinions.
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
OpenAI-O3-mini-medium
100.0%
#6
OpenAI-O4-mini-medium
100.0%
#7
GPT-OSS-120B
100.0%
#8
Claude-4.0-Sonnet-Thinking
100.0%
#9
OpenAI-O1-1217
100.0%
#10
OpenAI-O3-medium
100.0%
#11
Claude-3.7-Sonnet-Thinking
96.7%
#12
Claude-3.7-Sonnet
96.7%
#13
Claude-4.0-Opus-Thinking
96.7%
#14
Claude-4.0-Opus
96.7%
#15
Gemini-2.5-Pro-Thinking
96.7%
#16
GPT-4.5
96.7%
#17
Gemini-2.5-Pro-0325
96.7%
#18
Granite-3.0
96.7%
#19
Kimi-K2-Instruct
96.7%
#20
Gemini-2.0-Pro-0121
90.0%
#21
Gemini-2.5-Flash
90.0%
#22
Gemini-2.0-Flash
90.0%
#23
GPT-4-mini
90.0%
#24
GPT-4.1
90.0%
#25
Llama-4-Maverick-17B
86.7%
#26
Gemini-1.5-Pro
76.7%
#27
Llama-3.1-405B
76.7%
#28
GPT-4o-0513
73.3%
#29
DeepSeek-R1
71.7%
#30
Qwen-3
63.3%
#31
Qwen-3-Thinking
63.3%
#32
GPT-3.5-Turbo
56.7%
#33
DeepSeek-V3
43.3%
#34
Grok-3-Mini-Beta
40.0%
#35
Grok-3-Beta
33.3%
#36
Grok-4-Thinking
26.7%
#37
OpenAI-O1-mini
26.7%
#38
Grok-4
26.7%
#39
Cohere-Command-R-Plus
20.0%
#40
Mistral-Large-2
16.7%