Tax Agent Bench

Research-grade US tax questions answered by an agent. Run by Vals AI.

CategoryAgents
MetricAccuracy (%)
SourceVals AI (evaluator)
LicenseNone stated
Benchmark pagewww.vals.ai
StatusScored
Weight0.8 within Agents
FitDiscrimination 1.01, noise 0.44

Results

#ModelProviderResultDate
1 Gemini 4 Carbon High Google 51.8% 2026-11-09
2 Claude Fable 5.1 Max Anthropic 50.7% 2026-11-09
3 Gemini 4 Argon High Google 49.2% 2026-11-09
4 Claude Opus 5 Max Anthropic 47.1% 2026-11-09
5 Claude Opus 5.5 Max Anthropic 44.5% 2026-11-09
6 DeepSeek V4.1 Flash 1017 Max DeepSeek 44.4% 2026-11-09
7 GLM-5.3 Max Zhipu AI 43.3% 2026-11-09
8 Muse Wonder Full Max Meta 43.0% 2026-11-09
9 Muse Spark 1.3 Max Meta 42.4% 2026-11-09
10 Claude Sonnet 5.5 Max Anthropic 42.0% 2026-11-09
11 Inkling 2 Small Thinking Machines 38.9% 2026-11-09
12 Claude Fable 5 Max Anthropic 38.3% 2026-11-09
13 Muse Ember Max Meta 38.2% 2026-11-09
14 Muse Spark 1.3 Extra High Meta 37.9% 2026-11-09
15 Muse Wonder Max Meta 36.3% 2026-11-09
16 Grok 4.7 Extra High xAI 35.8% 2026-11-09
17 Hy4 Tencent 34.4% 2026-11-09
18 Grok 4.6 High xAI 33.7% 2026-11-09
19 Kimi K3 Max Moonshot AI 33.4% 2026-11-09
20 Muse Spark 1.2 Extra High Meta 32.8% 2026-11-09
21 Hy4 preview Tencent 32.6% 2026-11-09
22 MiMo-V2.6-Pro Xiaomi 32.4% 2026-11-09
23 MiniMax-M3.1-Flash Max MiniMax 32.1% 2026-11-09
24 Gemini 3.8 Flash High Google 32.1% 2026-11-09
25 Qwen3.8 Max Alibaba 32.1% 2026-11-09
26 Claude Fable 5.2 High Anthropic 32.1% 2026-11-09
27 Claude Haiku 5.5 Max Anthropic 31.9% 2026-11-09
28 GLM-5.3-Flash Max Zhipu AI 31.6% 2026-11-09
29 GPT-5.6 Sol Max OpenAI 31.1% 2026-11-09
30 Ember-1 Fireworks 31.1% 2026-11-09
31 Qwen3.8 27B Extra High Alibaba 30.9% 2026-11-09
32 Mistral Large 4 High Mistral 30.4% 2026-11-09
33 GPT-5.6 Terra Max OpenAI 30.0% 2026-11-09
34 Claude Opus 4.8 Max Anthropic 30.0% 2026-11-09
35 Claude Sonnet 5 Max Anthropic 29.4% 2026-11-09
36 MiMo-V2.6-Flash Xiaomi 28.5% 2026-11-09
37 DeepSeek V4 Flash 0731 High DeepSeek 28.1% 2026-11-09
38 Step 5 Preview StepFun 27.3% 2026-11-09
39 DeepSeek V4 Pro 0813 Max DeepSeek 26.8% 2026-11-09
40 DeepSeek V4.1 Flash Max DeepSeek 26.6% 2026-11-09
41 Muse Spark Open Extra High Meta 26.3% 2026-11-09
42 Claude Sonnet 4.6 Max Anthropic 25.8% 2026-11-09
43 DeepSeek V4 Pro Max DeepSeek 25.2% 2026-11-09
44 Kimi K3.1 Pro Max Moonshot AI 24.9% 2026-11-09
45 Grok 4.5 High xAI 24.9% 2026-11-09
46 GPT-5.6 Luna Max OpenAI 24.8% 2026-11-09
47 Claude Opus 4.7 Max Anthropic 23.7% 2026-11-09
48 Gemini 3.7 Flash High Google 22.8% 2026-11-09
49 GPT-5.5 Extra High OpenAI 21.8% 2026-11-09
50 GPT-6.1 Sol Max OpenAI 21.8% 2026-11-09
51 Gemini 3.5 Flash High Google 21.7% 2026-11-09
52 GPT-6 Astra Max OpenAI 20.7% 2026-11-09
53 MiniMax-M3 MiniMax 20.6% 2026-11-09
54 GPT-6 Luna Max OpenAI 19.9% 2026-11-09
55 Mercury 3 High Inception 18.2% 2026-11-09
56 Gemini 3.6 Flash High Google 18.0% 2026-11-09
57 GPT-6.1 Terra Max OpenAI 17.1% 2026-11-09
58 Qwen3.7 Max Alibaba 15.8% 2026-11-09
59 Inkling Small Thinking Machines 15.4% 2026-11-09
60 Inkling Thinking Machines 15.3% 2026-11-09
61 GPT-6 Sol Max OpenAI 15.0% 2026-11-09
62 Kimi K2.6 Moonshot AI 12.4% 2026-11-09
63 Gemini 3 Flash Preview High Google 11.8% 2026-11-09
64 Nemotron 3.5 Super VL 120B A12B High NVIDIA 11.6% 2026-11-09
65 MiMo-V2.5-Pro Xiaomi 11.5% 2026-11-09
66 Qwen3.7 Plus Alibaba 10.5% 2026-11-09
67 Gemini 3.1 Pro Preview High Google 9.3% 2026-11-09
68 Beam 1.0 Reflection 8.8% 2026-11-09
69 GPT-5.4 mini Extra High OpenAI 8.8% 2026-11-09
70 MiMo-V2.5 Xiaomi 7.8% 2026-11-09
71 Qwen3.6 Plus Alibaba 7.7% 2026-11-09
72 Gemini 3.5 Flash Lite High Google 7.3% 2026-11-09
73 Grok 4.20 (Reasoning) xAI 7.0% 2026-11-09
74 MiniMax-M2.7 MiniMax 6.9% 2026-11-09
75 Grok 4.3 High xAI 6.2% 2026-11-09
76 Claude Haiku 4.5 Thinking Anthropic 5.9% 2026-11-09
77 GPT-5.4 nano High OpenAI 5.0% 2026-11-09
78 Mistral Medium 3.5 High Mistral 4.5% 2026-11-09
79 Gemini 3.1 Flash Lite Preview High Google 3.1% 2026-11-09
80 Mercury 2.5 High Inception 2.5% 2026-11-09