BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases12

TaxEval

US tax questions checked by professionals. Run by Vals AI.

As of 24 Oct 2026, Muse Spark 1.2 leads TaxEval on BenchLeader with 80.4%, ahead of Muse Spark 1.1 at 79.7%, across 138 model configurations with a published result.

Published by
Vals AI
Category
Knowledge
Index weight
0.5
Models
138
Data as of
24 Oct 2026
Newest result
1 Sept 2026

Vals AI (vals.ai).

What the test looks like

TaxEval poses US tax questions, from filing rules to computed liabilities, checked by tax professionals.

How it is scored

Percent correct, run by Vals AI.

What to keep in mind

US-specific and dependent on the tax year of the questions; rules change annually.

Top 15
  1. 1Muse Spark 1.2 (xhigh)80.4%
  2. 2Muse Spark 1.1 (xhigh)79.7%
  3. 3Muse Spark77.7%
  4. 4Claude Sonnet 4.677.1%
  5. 5Claude Fable 576.9%
  6. 6GPT-5.6 Terra (xhigh)76.2%
  7. 7GPT-5.6 Luna (max)76.2%
  8. 8Claude Fable 5.176.0%
  9. 9Claude Opus 4.6 (thinking)76.0%
  10. 10Grok 375.9%
  11. 11GPT-5.2 (xhigh)75.8%
  12. 12Kimi K375.7%
  13. 13Grok 4 Fast (thinking)75.7%
  14. 14Claude Opus 4.875.6%
  15. 15Claude Sonnet 575.6%
138 of 138
#
1Muse Spark 1.2xhigh80.4%63.22026-09-01
2Muse Spark 1.1xhigh79.7%61.62026-09-01
3Muse Spark77.7%64.32026-09-01
4Claude Sonnet 4.677.1%57.12026-09-01
5Claude Fable 576.9%67.12026-09-01
6GPT-5.6 Terraxhigh76.2%62.82026-09-01
7GPT-5.6 Lunamax76.2%59.92026-09-01
8Claude Fable 5.176.0%66.42026-09-01
9Claude Opus 4.6thinking76.0%62.52026-09-01
10Grok 375.9%51.52026-09-01
11GPT-5.2xhigh75.8%60.72026-09-01
12Kimi K375.7%59.12026-09-01
13Grok 4 Fastthinking75.7%51.42026-09-01
14Claude Opus 4.875.6%60.12026-09-01
15Claude Sonnet 575.6%56.12026-09-01
16GLM 5.3 Flashmax75.6%54.22026-09-01
17Qwen3.8 Maxmax75.5%63.32026-09-01
18Inkling Small75.5%54.82026-09-01
19Qwen3.7max75.3%60.62026-09-01
20Claude Opus 4.775.3%63.32026-09-01
21GPT-5 minihigh75.2%54.12026-09-01
22Claude Opus 575.1%65.52026-09-01
23GPT-4.1high75.1%46.72026-09-01
24GPT-5.5xhigh75.0%66.12026-09-01
25Gemini 3.6 Flashhigh74.9%62.02026-09-01
26Claude Opus 4.5thinking74.9%59.92026-09-01
27GPT-5.1high74.9%57.72026-09-01
28GPT-5.6 Solmax74.8%68.02026-09-01
29o4-minihigh74.8%50.12026-09-01
30Gemini 3.7 Flashhigh74.7%63.52026-09-01
31Qwen3.6 Plus74.7%58.82026-09-01
32Kimi K2.674.7%55.42026-09-01
33o3high74.6%53.92026-09-01
34GPT-4ohigh74.5%41.72026-09-01
35Gemini 3.8 Flashhigh74.5%64.42026-09-01
36Gemini 3.5 Flashhigh74.4%63.12026-09-01
37Claude Opus 4.574.3%57.92026-09-01
38o1high74.3%48.32026-09-01
39Kimi K2.5thinking74.2%57.62026-09-01
40Grok 4.20thinking74.1%58.62026-09-01
41Claude 3.7 Sonnetthinking74.0%51.02026-09-01
42GPT-5.4xhigh74.0%64.72026-09-01
43Qwen3 Maxmax74.0%52.82026-09-01
44Gemini 3 Flashhigh73.9%58.32026-09-01
45MiMo-V2.5-Pro73.8%54.92026-09-01
46Claude Opus 4.1thinking73.7%56.42026-09-01
47GPT-5high73.4%57.92026-09-01
48GLM 5.273.3%52.02026-09-01
49Claude Sonnet 4.5thinking73.3%52.62026-09-01
50Grok 4.1 Fastthinking73.1%54.72026-09-01
51Nemotron 3 Ultra 550B A55B73.1%55.52026-09-01
52DeepSeek V4 Promax73.1%60.82026-09-01
53Mistral Large 373.1%45.42026-09-01
54Grok 3 Mini Fastlow73.0%50.72026-09-01
55Gemini 2.5 Pro72.9%53.92026-09-01
56Gemini 3.1 Prohigh72.9%60.32026-09-01
57MiniMax M372.7%56.42026-09-01
58Gemini 2.5 Flash 09 202572.7%46.92026-09-01
59Gemini 3.5 Flash Litehigh72.6%47.92026-09-01
60Gemini 3 Prohigh72.6%60.02026-09-01

More knowledge benchmarks · All LLM benchmarks

Cite as: BenchLeader, “TaxEval leaderboard”, data as of 24 Oct 2026, https://www.benchleader.com/benchmarks/vals_taxeval.

TaxEval: questions

What does TaxEval measure?
TaxEval poses US tax questions, from filing rules to computed liabilities, checked by tax professionals. Scores are reported in percent of tasks solved; higher is better.
Which AI model leads TaxEval?
Muse Spark 1.2 leads TaxEval with 80.4% as of 24 Oct 2026, ahead of Muse Spark 1.1 at 79.7%.
How many models have TaxEval results?
138 model configurations have a TaxEval result on BenchLeader, all taken from Vals AI.
Who runs TaxEval and how often is it updated?
TaxEval is published by Vals AI. BenchLeader re-reads the published results once a day and records the date each result was published.
Does TaxEval count toward the BenchLeader Index?
Yes. TaxEval contributes to the knowledge category of the BenchLeader Index, normalised so that 50 is the average of the evaluated models.