GPT AT THE POLLS

Political bias, quantified

31 models

Anthropic

Models are sorted like the Political Index when a finished analysis is available. Anything not finished is listed below as No analysis yet.

#
MODEL
ALIGNMENT
D%
R%
LEAN
01

Anthropic: Claude 3 Haiku

Anthropic
87
13
Strongly Left
02

Claude 3 Sonnet

Anthropic
86
14
Strongly Left
03

Claude 3 Opus

Anthropic
84
16
Strongly Left
04

Claude v2.0

Anthropic
82
18
Strongly Left
05

Anthropic: Claude 3.5 Haiku

Anthropic
80
20
Strongly Left
06

Anthropic: Claude 3.7 Sonnet (thinking)

Anthropic
80
20
Strongly Left
07

Anthropic: Claude Sonnet 4

Anthropic
78
22
Strongly Left
08

Anthropic: Claude Opus 4

Anthropic
77
23
Strongly Left
09

Anthropic: Claude 3.7 Sonnet

Anthropic
75
25
Strongly Left
10

Anthropic: Claude 3.5 Sonnet

Anthropic
74
26
Strongly Left
NO ANALYSIS YET21 models
Anthropic: Claude Haiku 4.5
No analysis yet
Anthropic: Claude Opus 4.1
No analysis yet
Anthropic: Claude Opus 4.5
No analysis yet
Anthropic: Claude Opus 4.6
No analysis yet
Anthropic: Claude Sonnet 4.5
No analysis yet
Anthropic: Claude Sonnet 4.6
No analysis yet
Claude 3 Haiku (self-moderated)
No analysis yet
Claude 3 Opus (self-moderated)
No analysis yet
Claude 3 Sonnet (self-moderated)
No analysis yet
Claude 3.5 Haiku (2024-10-22)
No analysis yet
Claude 3.5 Haiku (2024-10-22) (self-moderated)
No analysis yet
Claude 3.5 Haiku (self-moderated)
No analysis yet
Claude 3.5 Sonnet (2024-06-20)
No analysis yet
Claude 3.5 Sonnet (2024-06-20) (self-moderated)
No analysis yet
Claude 3.5 Sonnet (self-moderated)
No analysis yet
Claude 3.7 Sonnet (self-moderated)
No analysis yet
Claude v2
No analysis yet
Claude v2 (self-moderated)
No analysis yet
Claude v2.0 (self-moderated)
No analysis yet
Claude v2.1
No analysis yet
Claude v2.1 (self-moderated)
No analysis yet