GPT AT THE POLLS

Political bias, quantified

01 — Data Dashboard

Political Index

Latest published results for 137 AI models, based on up to 114 bill responses per model.

Models Tracked

137

Avg. Dem %

72%

Most Left

Anthropic: Claude 3 Haiku

Most Right

Meta: Llama 3.2 1B Instruct

Aggregate Alignment Distribution

Democrat-alignedRepublican-aligned

D — 72%

AVERAGE ACROSS ALL MODELS

R — 28%

#

MODEL

AGREEMENT

D%

R%

LEAN

01

Anthropic: Claude 3 Haiku

87

13

Strongly Left

02

Claude 3 Sonnet

86

14

Strongly Left

03

OpenAI: GPT-4.1

86

14

Strongly Left

04

R1 1776

86

14

Strongly Left

05

Google: Gemma 2 9B

85

15

Strongly Left

06

MoonshotAI: Kimi K2 0711

85

15

Strongly Left

07

Claude 3 Opus

84

16

Strongly Left

08

o1-preview

84

16

Strongly Left

09

OpenAI: GPT-3.5 Turbo 16k

83

17

Strongly Left

10

Claude v2.0

82

18

Strongly Left

11

NVIDIA: Llama 3.1 Nemotron Ultra 253B v1

82

18

Strongly Left

12

OpenAI: GPT-4o (2024-08-06)

82

18

Strongly Left

13

OpenAI: GPT-4o (2024-05-13)

82

18

Strongly Left

14

DeepSeek: R1

81

19

Strongly Left

15

Perplexity: Sonar

81

19

Strongly Left

16

Perplexity: Sonar Reasoning Pro

81

19

Strongly Left

17

Anthropic: Claude 3.5 Haiku

80

20

Strongly Left

18

Anthropic: Claude 3.7 Sonnet (thinking)

80

20

Strongly Left

19

Google: Gemma 3 4B

80

20

Strongly Left

20

DeepSeek: DeepSeek V3 0324

79

21

Strongly Left

21

GLM Z1 32B

79

21

Strongly Left

22

Llama 3.1 Sonar 70B Online

79

21

Strongly Left

23

Qwen: Qwen-Max

79

21

Strongly Left

24

Anthropic: Claude Sonnet 4

78

22

Strongly Left

25

Gemini 1.5 Pro

78

22

Strongly Left

26

Magnum v4 72B

78

22

Strongly Left

27

OpenAI: GPT-4o (2024-11-20)

78

22

Strongly Left

28

xAI: Grok 4

78

22

Strongly Left

29

Anthropic: Claude Opus 4

77

23

Strongly Left

30

Command R+

77

23

Strongly Left

31

DeepSeek: R1 Distill Llama 70B

77

23

Strongly Left

32

Google: Gemini 2.5 Pro Preview 05-06

77

23

Strongly Left

33

Google: Gemma 3 27B

77

23

Strongly Left

34

Grok 2 1212

77

23

Strongly Left

35

Qwen: Qwen3 8B

77

23

Strongly Left

36

Sonar Reasoning

77

23

Strongly Left

37

AionLabs: Aion-1.0

75

25

Strongly Left

38

Google: Gemini 2.5 Pro Preview 06-05

75

25

Strongly Left

39

Google: Gemma 2 27B

75

25

Strongly Left

40

Google: Gemma 3 12B

75

25

Strongly Left

41

Grok Beta

75

25

Strongly Left

42

Llama 3.3 Nemotron Super 49B v1

75

25

Strongly Left

43

NVIDIA: Llama 3.1 Nemotron 70B Instruct

75

25

Strongly Left

44

Perplexity: Sonar Pro

75

25

Strongly Left

45

Qwen: QwQ 32B

75

25

Strongly Left

46

Anthropic: Claude 3.7 Sonnet

75

25

Strongly Left

47

Codex Mini

75

25

Strongly Left

48

DeepSeek Prover V2

75

25

Strongly Left

49

DeepSeek: R1 0528 (free)

75

25

Strongly Left

50

Google: Gemini 2.5 Pro

75

25

Strongly Left

51

Google: Gemma 3n 4B (free)

75

25

Strongly Left

52

Meta: Llama 3 70B Instruct

75

25

Strongly Left

53

OpenAI: GPT-4 Turbo

75

25

Strongly Left

54

OpenAI: o4 Mini

75

25

Strongly Left

55

Qwen: Qwen3 14B

75

25

Strongly Left

56

Anthropic: Claude 3.5 Sonnet

74

26

Strongly Left

57

Magistral Small 2506

74

26

Strongly Left

58

Meta: Llama 3.1 70B Instruct

74

26

Strongly Left

59

Qwen: Qwen3 235B A22B

74

26

Strongly Left

60

AionLabs: Aion-1.0-Mini

73

27

Strongly Left

61

Amazon: Nova Lite 1.0

73

27

Strongly Left

62

Baidu: ERNIE 4.5 300B A47B

73

27

Strongly Left

63

DeepSeek: DeepSeek V3

73

27

Strongly Left

64

EVA Qwen2.5 72B

73

27

Strongly Left

65

Meta: Llama 3.1 405B Instruct

73

27

Strongly Left

66

Meta: Llama 3.3 70B Instruct

73

27

Strongly Left

67

Mistral: Saba

73

27

Strongly Left

68

OpenAI: o1

73

27

Strongly Left

69

Amazon: Nova Micro 1.0

72

28

Strongly Left

70

DeepSeek: R1 Distill Qwen 32B

72

28

Strongly Left

71

Goliath 120B

72

28

Strongly Left

72

Magistral Medium 2506 (thinking)

72

28

Strongly Left

73

Mistral: Mistral Small 3.1 24B

72

28

Strongly Left

74

Mistral: Mistral Small 3.2 24B

72

28

Strongly Left

75

OpenAI: GPT-4o-mini (2024-07-18)

72

28

Strongly Left

76

Qwen: Qwen3 32B

72

28

Strongly Left

77

Qwen2.5 72B Instruct

72

28

Strongly Left

78

Amazon: Nova Pro 1.0

71

29

Strongly Left

79

DeepHermes 3 Mistral 24B Preview (free)

71

29

Strongly Left

80

Kimi Dev 72B (free)

71

29

Strongly Left

81

Mistral: Mistral Nemo

71

29

Strongly Left

82

OpenAI: o4 Mini High

71

29

Strongly Left

83

R1 Distill Llama 8B

71

29

Strongly Left

84

WizardLM-2 8x22B

71

29

Strongly Left

85

GLM 4 32B

70

30

Strongly Left

86

Mistral: Devstral Small 1.1

70

30

Strongly Left

87

OpenAI: GPT-4.1 Mini

70

30

Strongly Left

88

Qwen 2 72B Instruct

70

30

Strongly Left

89

xAI: Grok 3 Mini Beta

70

30

Strongly Left

90

DeepSeek: R1 0528

69

31

Strongly Left

91

Devstral Small 2505

69

31

Strongly Left

92

Mistral: Mistral Medium 3

69

31

Strongly Left

93

Hunyuan A13B Instruct (free)

68

32

Strongly Left

94

Meta: Llama 3.1 8B Instruct

68

32

Strongly Left

95

Microsoft: Phi 4

68

32

Strongly Left

96

Mistral: Devstral Medium

68

32

Strongly Left

97

Moonlight 16B A3B Instruct (free)

68

32

Strongly Left

98

Deepseek R1 0528 Qwen3 8B

68

32

Strongly Left

99

MiniMax: MiniMax M1

68

32

Strongly Left

100

xAI: Grok 3 Mini

68

32

Strongly Left

101

Deepseek R1 0528 Qwen3 8B (free)

67

33

Strongly Left

102

Gemini 2.5 Flash Preview 05-20 (thinking)

67

33

Strongly Left

103

GPT-4 32k (older v0314)

67

33

Strongly Left

104

Llama 3.3 8B Instruct (free)

67

33

Strongly Left

105

Magistral Medium 2506

67

33

Strongly Left

106

Mercury Coder Small Beta

67

33

Strongly Left

107

Meta: Llama 4 Maverick

67

33

Strongly Left

108

Mistral: Ministral 8B

67

33

Strongly Left

109

Sarvam-M (free)

67

33

Strongly Left

110

Typhoon2 70B Instruct

67

33

Strongly Left

111

MiniMax M1 (extended)

66

34

Strongly Left

112

xAI: Grok 3 Beta

66

34

Strongly Left

113

Command

65

35

Strongly Left

114

GPT-4.5 (Preview)

65

35

Strongly Left

115

Mistral Large

65

35

Strongly Left

116

OpenAI: o3 Mini

65

35

Strongly Left

117

xAI: Grok 3

65

35

Strongly Left

118

Phi-3 Medium 128K Instruct

64

36

Leaning Left

119

Bytedance: UI-TARS 72B (free)

63

37

Leaning Left

120

Google: Gemini 2.5 Flash

63

37

Leaning Left

121

Midnight Rose 70B

sophosympatheia

63

37

Leaning Left

122

OpenAI: o3 Mini High

63

37

Leaning Left

123

Phi 4 Reasoning Plus

62

38

Leaning Left

124

o1-mini (2024-09-12)

61

39

Leaning Left

125

Gemini 2.5 Flash Lite Preview 06-17

61

39

Leaning Left

126

Meta: Llama 4 Scout

61

39

Leaning Left

127

OpenAI: GPT-4.1 Nano

60

40

Leaning Left

128

Dolphin 2.9.2 Mixtral 8x22B 🐬

Cognitive Computations

59

41

Leaning Left

129

Qwen: Qwen3 30B A3B

58

42

Leaning Left

130

Lumimaid v0.2 70B

57

43

Leaning Left

131

Gemini 2.5 Flash Preview 05-20

56

44

Centrist

132

Google: Gemini 2.0 Flash

53

47

Centrist

133

Gemini 2.5 Flash Preview 04-17

52

48

Centrist

134

R1 Distill Qwen 1.5B

52

48

Centrist

135

Typhoon2 8B Instruct

50

50

Centrist

136

Dobby Mini Plus Llama 3.1 8B

46

54

Centrist

137

Meta: Llama 3.2 1B Instruct

46

54

Centrist

ALIGNMENT SCALE

Strongly Left

>=65%

Leaning Left

57%-64%

Centrist

44%-56%

Leaning Right

36%-43%

Strongly Right

<=35%

Agreement is measured by comparing each model's simulated vote on congressional bills with the recorded votes of Rep. Alexandria Ocasio-Cortez and Speaker Mike Johnson. These rates describe agreement with two reference legislators, not agreement with either party as a whole.