Inference
Vals AI

TaxEval v2

A Vals-created set of questions and responses to tax questions

ModelsModels with a result.136
Top resultBest result on this test.80.4%Muse Spark 1.2
Top-3 spreadPoints from first to third.2.7 pts
UpdatedDate the source changed the results.1 Sept 2026

Result and cost

020406080100$0.00003$0.0001$0.0003$0.001$0.003$0.01$0.03$0.1$0.3$1RESULTUS DOLLARS PER TASK · LOG SCALE

Results

137 results
#ModelResultThe score from the source.CostUS dollars to run one task.TimeTime to run one task.
01Muse Spark 1.280.4%$0.01448 s
02Muse Spark 1.179.7%$0.01533 s
03Muse Spark77.7%<$0.0158 s
04Claude Sonnet 4.677.1%$0.1282 min
05Claude Fable 576.9%$0.2257 s
06GPT-5.6 Terra76.2%$0.03637 s
06GPT-5.6 Luna76.2%$0.01277 s
08Claude Fable 5.176.0%$0.34786 s
08Claude Opus 4.6 (Adaptive)76.0%$0.07179 s
10Grok 3 [Beta]75.9%$0.01713 s
11GPT-5.275.8%$0.06660 s
12Grok 4 Fast (Reasoning)75.7%<$0.0110 s
13Claude Opus 4.875.6%$0.15970 s
13Claude Sonnet 575.6%$0.3023 min
15GLM-5.3-Flash75.6%<$0.0153 s
16Qwen3.8 Max75.6%$0.0764 min
17Inkling-Small75.5%$0.0184 min
18Qwen3.7 Max75.3%$0.02858 s
18Inkling75.3%$0.1033 min
20Claude Opus 4.775.3%$0.12562 s
21GPT-5 mini75.2%$0.01237 s
22Claude Opus 575.1%$0.11458 s
23GPT-4.175.1%<$0.017.9 s
24GPT-5.575.0%$0.062 min
25Gemini 3.6 Flash74.9%$0.0212 s
25GPT-5.174.9%<$0.0144 s
27Claude Opus 4.5 Thinking74.9%$0.21947 s
28o4-mini (high)74.8%<$0.0115 s
29GPT-5.6 Sol74.8%$0.13862 s
30Gemini 3.7 Flash74.7%$0.0197.8 s
30Qwen3.6 Plus74.7%$0.01171 s
32o374.6%$0.01324 s
33Gemini 3.8 Flash74.4%$0.0248.8 s
34Gemini 3.5 Flash74.4%$0.03820 s
35Claude Opus 4.574.3%$0.05915 s
36o174.3%$0.12219 s
37Kimi K2.5 Thinking74.2%$0.0142 min
38Grok 4.20 (Reasoning)74.1%$0.01713 s
39Claude 3.7 Sonnet Thinking74.0%$0.05143 s
40GPT-5.474.0%$0.15779 s
40Qwen3 Maxqwen3-max-preview74.0%<$0.0121 s
42Gemini 3 Flash Preview73.9%<$0.0112 s
43MiMo-V2.5-Pro73.8%<$0.0163 s
44Claude Opus 4.1 Thinking73.7%$0.09327 s
45Qwen3 Maxqwen3-max73.5%<$0.0129 s
46GPT-573.4%$0.05479 s
47GLM-5.273.3%$0.01565 s
48Claude Sonnet 4.5 Thinking73.3%$0.04148 s
49Grok 4.1 Fast (Reasoning)73.1%<$0.0131 s
50Grok 3 Mini Fast Low Reasoning73.0%<$0.0112 s
51Gemini 2.5 Pro Exp 03.2572.9%$0.01122 s
52Gemini 3.1 Pro Preview72.9%$0.03142 s
53GPT-4o72.8%<$0.015.9 s
54MiniMax M372.7%<$0.012 min
54Gemini 2.5 Flash (Preview)72.7%<$0.016.3 s
56Gemini 3.5 Flash-Lite72.6%<$0.018.8 s
57Gemini 3 Pro72.6%$0.03424 s
58DeepSeek V4 Pro 042372.6%$0.0333 min
59GLM-4.572.4%<$0.012 min
59Claude 3.7 Sonnet72.4%<$0.016.3 s
59Gemini 2.5 Flash Preview Thinking72.4%<$0.0111 s
62GLM-5.372.4%$0.0322 min
63DeepSeek-R172.3%$0.0123 min
64Qwen3.5 Flash72.2%<$0.0173 s
65Claude Sonnet 4 Thinking72.0%$0.02932 s
66Claude Opus 471.9%$0.03715 s
66GPT-4.1 mini71.9%<$0.015.9 s
68MiMo-V2.571.8%<$0.0126 s
69Gemini 3.1 Flash Lite Preview71.8%<$0.017.6 s
70Kimi K2 Thinking71.7%<$0.0161 s
71Grok 4.571.7%$0.0193 min
72GPT-OSS 120B71.6%<$0.0148 s
73Grok 4 Fast Non Reasoning71.6%<$0.014.3 s
74Claude Opus 4.171.5%$0.0433 s
75Qwen3.6-27B71.3%$0.032 min
76GPT-5.4 mini71.2%<$0.0145 s
77GLM-5.171.2%$0.01784 s
78Gemini 2.5 Flash Preview 04.1771.2%<$0.015.4 s
79Grok 3 Mini Fast High Reasoning71.1%<$0.0125 s
80Grok 4.671.1%$0.01748 s
81Qwen3.8-27B70.8%$0.0462 min
82Grok 4.370.8%$0.0147 min
83DeepSeek V4 Flash 042370.7%<$0.0172 s
84Ling 3.0 Flash70.6%<$0.017.5 s
84Qwen3 235B A22B70.6%<$0.012 min
86Gemini 2.5 Flash Preview 04.17 Thinking70.5%<$0.0115 s
87Mistral Medium70.3%<$0.0111 s
88Kimi K2 Instruct70.2%<$0.0113 s
89Claude 3.5 Sonnet70.2%<$0.015.9 s
90GLM-5 Thinking70.0%$0.0222 min
91Gemini 2.0 Flash Thinking Exp 01.2169.8%<$0.0113 s
92Claude Sonnet 469.6%<$0.0112 s
93DeepSeek V369.5%<$0.0128 s
94o3-mini69.4%$0.01874 s
95GLM-4.768.8%$0.0143 min
96Mistral Large68.4%<$0.0130 s
97MiniMax M2.568.2%<$0.0173 s
97DeepSeek V3.2 (Thinking)68.2%<$0.012 min
97Gemini 2.0 Pro Exp 02.0568.2%<$0.019.0 s
100Mistral Medium 3.568.0%$0.0792 min
101Gemini 2.0 Flash Exp67.7%<$0.017.5 s
102Claude Haiku 4.5 Thinking67.5%$0.02748 s
103GPT-5.4 nano67.4%<$0.0116 s
104GPT-5 nano67.4%<$0.0170 s
105Grok 267.0%<$0.0111 s
106MiniMax M2.766.6%<$0.0129 s
106Llama4 Maverick Instruct Basic66.6%<$0.0125 s
108MiniMax M2.166.4%<$0.012 min
109GLM-4.666.2%<$0.0164 s
110Gemini 2.5 Flash Lite Preview66.2%<$0.013.6 s
111Gemini 2.0 Flash 00165.3%<$0.015.7 s
112Grok 465.1%$0.0783 min
113Gemini 2.5 Flash Lite Preview Thinking64.7%<$0.013.9 s
114Llama 3.3 Nemotron Super 49B V1 42e84561 Thinking63.7%—33 s
115GPT-OSS 20B63.7%<$0.0149 s
116Magistral Medium61.9%$0.02546 s
117Grok 4.1 Fast Non Reasoning61.9%<$0.019.3 s
118Command A61.4%<$0.0111 s
119Jamba Large 1.660.9%<$0.0116 s
119Meta Llama 3.1 405B Instruct Turbo60.9%<$0.0124 s
121GPT-4.1 nano60.8%<$0.013.0 s
122GPT-4o mini60.5%<$0.018.9 s
123Magistral Small60.3%<$0.0116 s
124Llama 3.3 Nemotron Super 49B V1 42e8456160.2%—18 s
125Gemini 1.5 Pro 00259.5%<$0.019.9 s
126Llama 3.3 70B Instruct Turbo59.4%<$0.013.8 s
127Laguna XS.259.0%<$0.0127 s
128Jamba 1.5 Large58.2%<$0.0120 s
129Claude 3.5 Haiku57.4%<$0.015.0 s
130Meta Llama 3.1 70B Instruct Turbo56.2%<$0.014.3 s
131Llama 4 Scout 17B 16E Instruct55.2%<$0.017.0 s
132Mistral Small53.7%<$0.018.0 s
133Gemini 1.5 Flash 00248.2%<$0.013.7 s
134Jamba Mini 1.644.6%<$0.014.3 s
135Jamba 1.5 Mini41.9%<$0.014.8 s
136Meta Llama 3.1 8B Instruct Turbo32.3%<$0.012.5 s
137Laguna M.11.6%—4 min