Artificial Analysis

Artificial Analysis IFBench

An Artificial Analysis IFBench score.

In index
Instruction
Best in instruction
Source ↗

Models147

Top result82.9%MiniMax M3

Top-3 spread1.6 pts

Year2026

Result and price

89 of 147 results

Results

147 results

#ModelResult

01MiniMax M382.9%

02Nemotron 3 Ultra81.4%

03Grok 4.381.3%

04Qwen3.7 Max80.5%

05MiMo-V2.5-Pro79.9%

06Qwen3.5 397B (Reasoning)78.8%

06Qwen3.5 397B78.8%

08Qwen3.7 Plus78.0%

09GPT-5.2-Codex77.6%

10Gemini 3.1 Pro77.1%

11Qwen 3.6 Max (preview)76.6%

12DeepSeek V4 Pro 081376.5%

13Gemini 3.5 Flash76.3%

13GLM-5.176.3%

15Kimi K2.676.0%

16GPT-5.575.9%

16Muse Spark75.9%

16GPT-5.4 nano75.9%

19Qwen3.5-122B-A10B75.7%

19MiniMax M2.775.7%

21Qwen3.5-27B75.6%

21Gemma 4 31B75.6%

23GPT-5.3 Codex75.4%

23GPT-5.275.4%

25Qwen3.6 Plus75.2%

26GPT-5.473.9%

26Command A+Above fit73.9%

28Gemma 4 12BAbove fit73.5%

29GLM-5.273.3%

29GPT-5.4 mini73.3%

31GLM-5-Turbo73.2%

32GPT-5 (high)73.1%

33GPT-5.172.9%

34GPT-5.6 Sol72.7%

35Qwen3.5-35B-A3B72.5%

36Gemma 4 26B A4B72.4%

37GLM-572.3%

38o371.4%

39GPT-5.6 Terra71.2%

39Solar Pro 3Above fit71.2%

41GPT-5 (medium)70.6%

42Gemini 3 Pro70.4%

43o170.3%

44Kimi K2.5 (Reasoning)70.2%

44Kimi K2.570.2%

46GPT-5.1-Codex-Max70.0%

46GPT-5.1-Codex70.0%

48GPT-OSS 120B69.0%

49MiMo-V2-Pro68.8%

49Mistral Medium 3.5 128B68.8%

51o4-mini68.7%

52GLM-4.767.9%

53Qwen3.6-27B67.6%

54Step 3.7 Flash67.3%

55GPT-OSS 20B65.1%

56K-Exaone64.7%

57Qwen3.6-35B-A3B64.4%

58Claude Fable 563.5%

59Nemotron 3 Nano Omni 30B A3B63.2%

60Kimi K2.7 Code63.1%

61Claude Opus 4.862.2%

62GLM-5V-Turbo61.1%

62Nemotron 3.5 Bolt E4B61.1%

64Claude Opus 4.7 (Adaptive)58.6%

65Claude Opus 4.5 Thinking58.0%

66North Mini Code57.6%

67Ling 2.6 Flash57.4%

68Trinity-Large-Thinking56.3%

68Trinity-Large-Preview56.3%

70LFM2.5-8B-A1B55.6%

71Claude 4.1 Opus Thinking55.4%

72Gemini 3 Flash55.1%

73Grok 453.7%

73Claude 4 Opus Thinking53.7%

75MiMo-V2-Omni53.5%

76Claude Opus 4.6 (Adaptive)53.1%

77Grok 4.1 Fast (Reasoning)52.7%

78Qwen3.5-4B52.0%

79Grok 4 Fast (Reasoning)50.5%

80DeepSeek V3.249.0%

81Gemini 2.5 Pro48.7%

82Claude 3.7 Sonnet Thinking48.3%

83Mistral Small 448.2%

—Mistral Small 4 (Reasoning)48.2%

84Llama 3.3 70B47.1%

84Ultravox v0.6 Llama 3.3 70B47.1%

86Claude 4 Sonnet45.4%

87Claude Opus 4.644.6%

88Gemma 4 E4B44.2%

89Qwen3 Max44.1%

90Claude 3.7 Sonnet44.0%

91Claude Opus 4.743.6%

92Qwen3-Omni-30B-A3B-Thinking43.4%

93Claude 4 Opus43.3%

94Claude Opus 4.543.0%

94GPT-4.143.0%

94Llama 4 Maverick43.0%

97Claude 3.5 Haiku42.8%

98DeepSeek V3.1 (Reasoning)41.5%

98Kimi K241.5%

100Grok Code Fast 141.4%

101Claude Sonnet 4.641.2%

102DeepSeek V3 032441.0%

103Gemini 2.0 Flash40.2%

104MiMo-V2-Flash39.9%

105DeepSeek-R139.6%

106Llama 4 Scout39.5%

107Mistral Medium 339.3%

108Gemini 2.5 Flash39.0%

108Llama 3.1 405B39.0%

110QwQ-32B38.8%

111GPT-4.1 mini38.3%

112Nemotron Ultra 253B38.2%

113Nova Pro38.1%

114Gemma 4 E2B38.0%

115DeepSeek V3.137.8%

116GLM-4.5-Air37.6%

117Nemotron 3 Nano 30B37.5%

118GLM-4.636.7%

119Grok 4.1 Fast36.5%

120Mistral Large 336.2%

121Claude 3 Haiku36.1%

122DeepSeek V334.8%

123Devstral Small34.6%

124Sarvam 105B34.4%

125GPT-4o34.3%

126Solar Pro 233.7%

127Exaone 4.0 32B33.5%

128LFM2.5-VL-1.6B-Extract33.1%

129GPT-4.1 nano32.0%

130Gemma 3 27B31.8%

131Gemini 2.5 Flash-Lite31.5%

132Mistral Large 231.2%

132Qwen3-Omni-30B-A3B-Instruct31.2%

134GPT-4o mini31.0%

135Mistral Small 3.129.9%

136Llama 3.1 70B29.0%

137Sarvam 30B26.5%

138Granite-4.0-H-1B26.2%

139Exaone 4.0 1.2B25.3%

140Magistral Medium25.1%

141Magistral Small24.8%

142Phi-4-reasoning23.5%

142Phi-423.5%

144DeepSeek R1 Distill Qwen 32B22.9%

145Granite-4.0-H-350M17.6%

146Granite-4.0-350M15.9%