Vals AI

LegalBench

Evaluating language models on a wide range of open source legal reasoning tasks.

Not in index
Legal
Source ↗

Models148

Top result88.6%Claude Fable 5

Top-3 spread0.3 pts

Updated13 Oct 2026

Result and cost

75 of 149 results

Results

149 results

#ModelResultCostTime

01Claude Fable 588.6% ±0.3$0.0319.0 s

02Claude Fable 5.188.5% ±0.4$0.05212 s

03Gemini 4 Argon88.3% ±0.4$0.01613 s

04Muse Wonder Full87.4% ±0.4$0.02538 s

04Gemini 3.1 Pro Preview87.4% ±0.3$0.006810 s

06Gemini 3.7 Flash87.3% ±0.4$0.00412.2 s

07Gemini 3 Pro87.0% ±0.4$0.018.3 s

07Gemini 3.8 Flash87.0% ±0.4$0.00813.3 s

07Claude Opus 587.0% ±0.4$0.0124.4 s

07GPT-5.6 Sol87.0% ±0.4$0.0136.2 s

11Gemini 3 Flash Preview86.9% ±0.4$0.00214.7 s

12Gemini 3.6 Flash86.7% ±0.4$0.00483.0 s

13GPT-5.586.5% ±0.4$0.01518 s

14Grok 4.686.3% ±0.4$0.01527 s

15GLM Legal86.2% ±0.4$0.000146.0 s

15Kimi K386.2% ±0.4$0.00797.1 s

17GPT-5.486.0% ±0.4$0.02328 s

17GPT-586.0% ±0.4—19 s

17Grok 4.586.0% ±0.4$0.004568 s

20GPT-5.185.7% ±0.4$0.00487.0 s

21MiniMax M385.4% ±0.4$0.000918.2 s

22Claude Opus 4.6 (Adaptive)85.3% ±0.4$0.0043.8 s

22Muse Spark 1.285.3% ±0.4$0.006430 s

22Claude Opus 4.785.3% ±0.5$0.008327 s

25GPT-5.6 Terra85.1% ±0.5$0.00262.8 s

25Qwen3.5 Plus Thinking85.1% ±0.4$0.003730 s

27Muse Spark 1.185.0% ±0.4$0.008517 s

28Qwen3.7 Max84.9% ±0.4$0.00612 s

29GLM-5.384.8% ±0.4$0.006522 s

30Kimi K2.684.7% ±0.4$0.006550 s

31Claude Opus 4.5 Thinking84.6% ±0.4$0.0377.6 s

32Grok 4.384.5% ±0.4$0.00316 min

33GLM-5.184.4% ±0.4$0.002214 s

33Grok 4.784.4% ±0.5$0.01126 s

35Gemini 2.5 Pro Exp 03.2584.3% ±0.4—4.0 s

35Qwen3.5 Flash84.3% ±0.4$0.0007617 s

37Qwen3.6 Plus84.2% ±0.4$0.00319 s

37Muse Spark84.2% ±0.4$0.001234 s

39Claude Sonnet 4.5 Thinking84.1% ±0.4—13 s

39GLM-5.284.1% ±0.5$0.00165.8 s

39Gemini 3.5 Flash-Lite84.1% ±0.4$0.00131.7 s

39GLM-5 Thinking84.1% ±0.4$0.002316 s

43GPT-5.6 Luna84.0% ±0.4$0.000945.9 s

43MiniMax M2.784.0% ±0.4$0.000638.9 s

45GLM-5.3-Flash83.9% ±0.4$0.0002813 s

45Claude Sonnet 583.9% ±0.5$0.00584.9 s

47Gemini 2.5 Flash Preview 04.1783.8% ±0.4—0.5 s

47Gemini 3.1 Flash Lite Preview83.8% ±0.4$0.00032.7 s

47o383.8% ±0.4—6.0 s

47Hy4 preview83.8% ±0.4$0.0132 min

51Qwen3.8 Max83.6% ±0.4$0.0136 s

51Gemini 3.5 Flash83.6% ±0.8$0.007635 s

51Claude Opus 4.883.6% ±0.5$0.00913.1 s

54Claude Opus 4.183.5% ±0.4—2.8 s

55GLM-4.783.4% ±0.5$0.002331 s

56DeepSeek V4.1 Flash 101783.3% ±0.5$0.00146.5 s

57Grok 483.2% ±0.5—27 s

58Grok 3 Mini Fast High Reasoning83.1% ±0.4—5.1 s

58GPT-4.183.1% ±0.4—0.5 s

58Claude Opus 483.1% ±0.4—3.7 s

58Mercury 2.583.1% ±0.5$0.000793.9 s

62Inkling-Small83.0% ±0.5$0.001318 s

62Claude Sonnet 483.0% ±0.4—4.8 s

64Inkling82.9% ±0.5$0.01323 s

65Claude Opus 4.582.8% ±0.4$0.0132.4 s

65GPT-5.282.8% ±0.4$0.01817 s

67Gemini 2.5 Flash Preview Thinking82.6% ±0.4—2.0 s

67Grok 3 [Beta]82.6% ±0.4—0.7 s

69Claude 3.7 Sonnet Thinking82.5% ±0.5—6.5 s

69Gemini 2.5 Flash (Preview)82.5% ±0.4—1.2 s

69Grok 4.1 Fast (Reasoning)82.5% ±0.5$0.000388.6 s

72Qwen3.8-27B82.4% ±0.5$0.004748 s

72DeepSeek V4 Pro 081382.4% ±0.4$0.004120 s

74GPT-4o82.2% ±0.5—0.5 s

75Claude Sonnet 4.682.1% ±0.5$0.004222 s

75Claude Sonnet 4 Thinking82.1% ±0.5—11 s

77Gemini 2.5 Flash Lite Preview Thinking82.0% ±0.4—3.1 s

78Grok 3 Mini Fast Low Reasoning81.9% ±0.4—3.4 s

78Qwen3 Maxqwen3-max81.9% ±0.4—0.0 s

80GPT-5 mini81.8% ±0.5—9.4 s

81Kimi K2 Instruct81.5% ±0.4—1.2 s

82Claude Haiku 4.5 Thinking81.2% ±0.5—10 s

83DeepSeek V380.8% ±0.4—4.4 s

84Grok 4 Fast (Reasoning)80.6% ±0.5—3.8 s

85GPT-4 Turbo80.5% ±0.5—1.4 s

86o180.4% ±0.6—8.3 s

87Qwen3 Maxqwen3-max-preview80.3% ±0.4—1.9 s

87DeepSeek V4 Pro 042380.3% ±0.5$0.00483 min

89Kimi K2 Thinking80.2% ±0.5$0.002244 s

89Qwen3 235B-A22B80.2% ±0.4—12 s

91Claude 3.7 Sonnet80.0% ±0.5—1.2 s

91MiniMax M2.580.0% ±0.5$0.001316 s

93Command A79.7% ±0.4—0.8 s

93Ling 3.0 Flash79.7% ±0.5$0.000173.9 s

95GLM-4.679.6% ±0.5—30 s

96Qwen2.5 72B Instruct Turbo79.4% ±0.7—0.8 s

97o4-mini79.2% ±0.5—4.0 s

98Mistral Large 379.1% ±0.5$0.000441.7 s

98Grok 4.1 Fast Non Reasoning79.1% ±0.5$0.0000911.8 s

100Gemini 2.5 Flash Lite Preview79.0% ±0.4—0.5 s

101MiMo-V2.578.9% ±0.5$0.0002611 s

102Grok 4 Fast Non Reasoning78.4% ±0.5—0.9 s

102Gemini 2.0 Flash 00178.4% ±0.8—0.4 s

104GPT-4.1 mini78.0% ±0.4—0.5 s

105GPT-5.4 nano77.9% ±0.4$0.000112.7 s

106Llama4 Maverick Instruct Basic77.8% ±0.4—0.5 s

107Grok 4.20 (Reasoning)77.7% ±0.5$0.00617.6 s

107DeepSeek V3 032477.7% ±0.4—7.7 s

107DeepSeek V4 Flash 073177.7% ±0.5$0.000523.9 s

110Llama 3.3 70B Instruct Turbo77.2% ±0.8—0.5 s

111MiMo-V2.5-Pro77.1% ±0.6$0.0008527 s

112DeepSeek V3.2 (Thinking)76.1% ±0.5$0.0004531 s

113GPT-OSS 120B75.9% ±0.5—7.9 s

114GLM-4.575.6% ±0.5—16 s

115MiniMax M2.175.4% ±0.5$0.0009411 s

116Laguna M.175.1% ±0.6—42 s

117Jamba 1.5 Large74.2% ±0.6—0.8 s

118Llama 4 Scout 17B 16E Instruct72.0% ±0.5—0.4 s

118Jamba Large 1.672.0% ±0.4—0.8 s

120o3-mini71.5% ±0.7—12 s

121Laguna XS.271.0% ±0.6—12 s

122GPT-OSS 20B70.8% ±0.6—7.2 s

123Claude 3.5 Haiku70.3% ±0.6—1.1 s

124Gemini 1.0 Pro 00270.2% ±0.8—0.5 s

125Jamba Mini 1.669.7% ±0.5—0.3 s

126Qwen2.5 7B Instruct Turbo69.6% ±0.8—0.4 s

127Mistral Small 3.169.2% ±0.5—0.6 s

128Gemini 1.5 Pro 00269.1% ±0.8—0.9 s

129Ling 3.0 Flash Af Rc368.7% ±0.5$0.00026.3 s

130Command R Plus68.3% ±0.7—0.4 s

131Gemma 2 9B IT68.1% ±0.7—0.6 s

132Gemma 2 27B68.0% ±0.7—0.9 s

133DeepSeek-R167.3% ±0.5—19 s

134Jamba 1.5 Mini66.6% ±0.7—0.3 s

135DeepSeek V3.266.0% ±0.4$0.0000862.2 s

136GPT-3.5 Turbo64.4% ±0.6—0.5 s

137Gemini 1.5 Flash 00263.2% ±0.7—0.5 s

138Llama 2 70B Hf62.4% ±0.7—14 s

139Mistral Medium62.0% ±0.5—0.7 s

140Command A+61.4% ±0.6$0.03323 s

141GPT-4.1 nano61.1% ±0.6—0.4 s

142Mixtral 8x7B V0.155.8% ±0.7—14 s

143Magistral Medium54.8% ±0.6—12 s

144Mistral 7B V0.153.8% ±0.8—1.7 s

145Llama 2 13B53.7% ±0.7—1.7 s

146Llama 2 7B51.9% ±0.6—1.0 s

147GPT-5 nano50.1% ±0.5—22 s

148Magistral Small40.0% ±0.5—6.7 s

149Command R33.0% ±0.7—0.2 s