Artificial Analysis

Artificial Analysis IFBench

An Artificial Analysis IFBench score.

In index
Instruction
Source ↗

Models129

Top result82.9%MiniMax M3

Top-3 spread1.6 pts

Year2026

Result and price

Results

129 results

#ModelResult

01

MiniMax M382.9%

02

Nemotron 3 Ultra81.4%

03

Grok 4.381.3%

04

Qwen3.7 Max80.5%

05

MiMo-V2.5-Pro79.9%

06

Qwen3.7 Plus78.0%

07

GPT-5.2-Codex77.6%

08

Gemini 3.1 Pro77.1%

09

Qwen 3.6 Max (preview)76.6%

10

DeepSeek V4 Pro 081376.5%

11

Gemini 3.5 Flash76.3%

11

GLM-5.176.3%

13MAKimi K2.676.0%

14

GPT-5.575.9%

14

Muse Spark75.9%

14

GPT-5.4 nano75.9%

17

Qwen3.5-122B-A10B75.7%

17

MiniMax M2.775.7%

19

Qwen3.5-27B75.6%

19

Gemma 4 31B75.6%

21

GPT-5.3 Codex75.4%

21

GPT-5.275.4%

23

Qwen3.6 Plus75.2%

24

GPT-5.473.9%

24COCommand A+73.9%

26

Gemma 4 12B73.5%

27

GLM-5.273.3%

27

GPT-5.4 mini73.3%

29

GLM-5-Turbo73.2%

30

GPT-5 (high)73.1%

31

GPT-5.172.9%

32

GPT-5.6 Sol72.7%

33

Qwen3.5-35B-A3B72.5%

34

Gemma 4 26B A4B72.4%

35

GLM-572.3%

36

Nemotron 3 Super 100B71.5%

37

o371.4%

38

GPT-5.6 Terra71.2%

38UPSolar Pro 371.2%

40

Nemotron 3 Nano 30B71.1%

41

GPT-5 (medium)70.6%

42

Gemini 3 Pro70.4%

43

o170.3%

44MAKimi K2.5 (Reasoning)70.2%

44MAKimi K2.570.2%

46

GPT-5.1-Codex-Max70.0%

46

GPT-5.1-Codex70.0%

48

GPT-OSS 120B69.0%

49

MiMo-V2-Pro68.8%

49

Mistral Medium 3.5 128B68.8%

51

GLM-4.767.9%

52

Qwen3.6-27B67.6%

53STStep 3.7 Flash67.3%

54

GPT-OSS 20B65.1%

55LAK-Exaone64.7%

56

Qwen3.6-35B-A3B64.4%

57

Claude Fable 563.5%

58

Nemotron 3 Nano Omni 30B A3B63.2%

59MAKimi K2.7 Code63.1%

60

Claude Opus 4.862.2%

61

GLM-5V-Turbo61.1%

62

Claude Opus 4.7 (Adaptive)58.6%

63

Claude Opus 4.5 Thinking58.0%

64CONorth Mini Code57.6%

65INLing 2.6 Flash57.4%

66AATrinity-Large-Thinking56.3%

66AATrinity-Large-Preview56.3%

68LILFM2.5-8B-A1B55.6%

69

Claude 4.1 Opus Thinking55.4%

70

Gemini 3 Flash55.1%

71

Grok 453.7%

72

MiMo-V2-Omni53.5%

73

Claude Opus 4.6 (Adaptive)53.1%

74

Grok 4.1 Fast (Reasoning)52.7%

75

Qwen3.5 397B51.6%

75

Qwen3.5 397B (Reasoning)51.6%

77

Grok 4 Fast (Reasoning)50.5%

78

DeepSeek V3.249.0%

79

Gemini 2.5 Pro48.7%

80

Mistral Small 448.2%

—

Mistral Small 4 (Reasoning)48.2%

81FAUltravox v0.6 Llama 3.3 70B47.1%

82

Claude 4 Sonnet45.4%

83

Claude Opus 4.644.6%

84

Gemma 4 E4B44.2%

85

Qwen3 Max44.1%

86

Claude Opus 4.743.6%

87

Qwen3-Omni-30B-A3B-Thinking43.4%

88

Claude Opus 4.543.0%

88

GPT-4.143.0%

88

Llama 4 Maverick43.0%

91MAKimi K241.5%

91

DeepSeek V3.1 (Reasoning)41.5%

93

Grok Code Fast 141.4%

94

Claude Sonnet 4.641.2%

95

MiMo-V2-Flash39.9%

96

DeepSeek-R139.6%

97

Llama 4 Scout39.5%

98

Mistral Medium 339.3%

99

Gemini 2.5 Flash39.0%

99

Llama 3.1 405B39.0%

101

GPT-4.1 mini38.3%

102

Nemotron Ultra 253B38.2%

103AMNova Pro38.1%

104

Gemma 4 E2B38.0%

105

DeepSeek V3.137.8%

106

GLM-4.5-Air37.6%

107

GLM-4.636.7%

108

Grok 4.1 Fast36.5%

109

Mistral Large 336.2%

110

Claude 3 Haiku36.1%

111

DeepSeek V334.8%

112SASarvam 105B34.4%

113

GPT-4o34.3%

114UPSolar Pro 233.7%

115LAExaone 4.0 32B33.5%

116LILFM2.5-VL-1.6B-Extract33.1%

117

GPT-4.1 nano32.0%

118

Gemma 3 27B31.8%

119

Mistral Large 231.2%

119

Qwen3-Omni-30B-A3B-Instruct31.2%

121

GPT-4o mini31.0%

122SASarvam 30B26.5%

123IBGranite-4.0-H-1B26.2%

124LAExaone 4.0 1.2B25.3%

125

Phi-423.5%

126

DeepSeek R1 Distill Qwen 32B22.9%

127IBGranite-4.0-H-350M17.6%

128IBGranite-4.0-350M15.9%