Crosshair
Benchmarks
Tax & Accountinghigher is better

TaxEval

Vals AI TaxEval v2 — 1,500+ expert-written tax questions, scored on overall accuracy. Independent, in-house-run.

Benchmark source
Domain
Tax & Accounting
Metric
%
Orientation
Higher is better
Results
25

Ranking

#ModelScoreSourceStatus
1Muse Spark
Meta AI
77.7%Vals AI — TaxEval v23rd-partyunverified
2Claude Sonnet 4.6
Anthropic
77.1%Vals AI — TaxEval v23rd-partyunverified
3Claude Opus 4.6
Anthropic
76%Vals AI — TaxEval v23rd-partyunverified
4GPT-5.2
OpenAI
75.8%Vals AI — TaxEval v23rd-partyunverified
5Claude Opus 4.8
Anthropic
75.6%Vals AI — TaxEval v23rd-partyunverified
6Claude Opus 4.7
Anthropic
75.3%Vals AI — TaxEval v23rd-partyunverified
7Qwen3.7 Max
Alibaba Qwen
75.3%Vals AI — TaxEval v23rd-partyunverified
8GPT-5.5
OpenAI
75%Vals AI — TaxEval v23rd-partyunverified
9Kimi K2.6
Moonshot AI
74.7%Vals AI — TaxEval v23rd-partyunverified
10Gemini 3.5 Flash
Google DeepMind
74.4%Vals AI — TaxEval v23rd-partyunverified
11Grok 4.20
xAI
74.1%Vals AI — TaxEval v23rd-partyunverified
12GPT-5.4
OpenAI
74%Vals AI — TaxEval v23rd-partyunverified
13Gemini 3 Flash
Google DeepMind
73.9%Vals AI — TaxEval v23rd-partyunverified
14Mistral Large 3
Mistral AI
73.1%Vals AI — TaxEval v23rd-partyunverified
15Gemini 3.1 Pro
Google DeepMind
72.9%Vals AI — TaxEval v23rd-partyunverified
16Gemini 3 Pro
Google DeepMind
72.6%Vals AI — TaxEval v23rd-partyunverified
17DeepSeek V4-Pro
DeepSeek
72.1%Vals AI — TaxEval v23rd-partyunverified
18Kimi K2 Thinking
Moonshot AI
71.7%Vals AI — TaxEval v23rd-partyunverified
19Qwen3.6-27B
Alibaba Qwen
71.3%Vals AI — TaxEval v23rd-partyunverified
20GLM-5.1
Z.ai (Zhipu)
71.2%Vals AI — TaxEval v23rd-partyunverified
21Grok 4.3
xAI
70.8%Vals AI — TaxEval v23rd-partyunverified
22DeepSeek V3.2
DeepSeek
68.2%Vals AI — TaxEval v23rd-partyunverified
23Claude Haiku 4.5
Anthropic
67.5%Vals AI — TaxEval v23rd-partyunverified
24Llama 4 Maverick
Meta AI
66.6%Vals AI — TaxEval v23rd-partyunverified
25Llama 4 Scout
Meta AI
55.2%Vals AI — TaxEval v23rd-partyunverified