Benchmark Details - Model Trust Scores

Chatbot Arena (Win Rate)

Measures the model's performance in head-to-head comparisons with other models in the Chatbot Arena, where human judges evaluate the quality of responses.

Source:

Model Performance

Gemini 3.1 Pro Preview

100.7%

Claude Opus 4.6 (Thinking)

100.4%

Grok 4.20

99.4%

Grok 4.20 (Reasoning)

99.4%

Claude Opus 4.6

98.6%

Claude Opus 4.5 (Thinking)

95.1%

GPT-5.4

95.0%

GPT-5.2

94.9%

Claude Opus 4.5

94.6%

#10

Gemini 2.5 Pro (Jun 2025)

94.3%

#11

Claude Sonnet 4.6

94.3%

#12

Gemini 2.5 Pro

94.3%

#13

Gemini 2.5 Pro (Thinking)

94.3%

#14

GLM-5

93.1%

#15

Kimi K2.5

93.0%

#16

GLM-4.7

92.1%

#17

GPT-5 (Thinking)

92.0%

#18

Grok 4 (Thinking)

91.7%

#19

Grok 4

91.7%

#20

Claude Sonnet 4.5 (Thinking)

90.3%

#21

Qwen 3 Max Preview

90.1%

#22

Mistral Large 3

89.7%

#23

Claude 4.1 Opus (Thinking)

89.3%

#24

OpenAI o3 (Medium Effort)

89.1%

#25

Grok 3

89.0%

#26

DeepSeek V3.2 (Thinking)

88.9%

#27

Claude Sonnet 4.5

88.6%

#28

Claude 4.1 Opus

88.4%

#29

DeepSeek V3.2

88.3%

#30

Qwen 3 (Thinking)

88.0%

#31

Qwen 3

88.0%

#32

GPT-4.5

87.9%

#33

GPT-5

87.6%

#34

Gemini 2.5 Flash

87.4%

#35

Gemini 2.0 Pro

85.4%

#36

Gemini 2.0 Flash Thinking (Jan 2025)

85.3%

#37

Gemini 2.0 Flash (Thinking)

85.3%

#38

GPT-4.1

83.0%

#39

Kimi K2

82.9%

#40

Claude Haiku 4.5

82.6%

#41

DeepSeek V3 (Mar 2025)

82.4%

#42

DeepSeek V3

82.4%

#43

Claude 4.0 Opus (Thinking)

82.3%

#44

GPT-5 Mini (Thinking)

82.1%

#45

GPT-5 Mini

82.1%

#46

DeepSeek R1

81.9%

#47

Gemini 2.0 Flash

81.4%

#48

Mistral Medium 3.1

81.3%

#49

GPT OSS 120B

81.1%

#50

Claude 4.0 Opus

80.9%

#51

Grok 3 Mini

80.9%

#52

OpenAI o1

80.9%

#53

OpenAI o4 Mini (Medium Effort)

80.3%

#54

OpenAI o4 Mini (High Effort)

80.3%

#55

Claude 4.0 Sonnet (Thinking)

78.7%

#56

Llama 3.3 70B

77.1%

#57

OpenAI o3 Mini (High Effort)

76.9%

#58

GPT-4o Mini

76.9%

#59

Claude 4.0 Sonnet

76.4%

#60

GPT-5 Nano

76.1%

#61

GPT-5 Nano (Thinking)

76.1%

#62

Cohere Command A

75.3%

#63

Llama 3.1 405B

74.4%

#64

Gemini 1.5 Pro

74.3%

#65

OpenAI o3 Mini (Medium Effort)

74.0%

#66

OpenAI o1 Mini

74.0%

#67

Claude 3.7 Sonnet (Thinking)

73.7%

#68

GPT-4o

71.7%

#69

Claude 3.7 Sonnet

71.6%

#70

Claude 3.5 Sonnet

71.3%

#71

Llama 4 Maverick 17B

70.3%

#72

Mistral Large 2

67.0%

#73

Claude 3 Opus

66.4%

#74

Claude 3.5 Haiku

65.1%

#75

Cohere Command R+

61.9%

#76

Claude 3 Sonnet

60.4%

#77

Phi-4

60.3%

#78

GPT-3.5 Turbo

49.3%

#79

Llama 2 70B

46.0%

#80

Granite 3.0

44.0%

#81

Llama 2 13B

41.9%

Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

Chatbot Arena (Win Rate)

Model Performance