BenchLeader
Menu
Best LLMsChoose a modelCoding plansLeaderboardsModelsBenchmarksProvidersReleases7

Tax Agent Bench leaderboard

As of 6 Nov 2026, the top score on Tax Agent Bench is 51.8% by Gemini 4 Carbon (Google), ahead of Qwen3.8 Max Internal at 49.7%, across 79 model configurations with a published result (Vals AI, result dated 5 Nov 2026). Corporate tax questions that require researching the code and regulations with tools, graded by tax professionals. Tax Agent Bench is shown for reference only and does not count toward the BenchLeader Index.

Key facts (as of 6 Nov 2026)

  • Gemini 4 Carbon holds the top Tax Agent Bench score at 51.8% (Vals AI, 5 Nov 2026).
  • Qwen3.8 Max Internal is second on Tax Agent Bench at 49.7% (Vals AI, 5 Nov 2026).
  • Tax Agent Bench: Agentic US corporate tax research. Run by Vals AI (Vals AI).
  • 79 model configurations have a published Tax Agent Bench result on BenchLeader (BenchLeader, 6 Nov 2026).
  • Tax Agent Bench does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).

Agentic US corporate tax research. Run by Vals AI.

Published by
Vals AI
Category
Agents & tools
Index weight
Reference only
Models
79
Data as of
6 Nov 2026

Vals AI (vals.ai).

What the test looks like

Corporate tax questions that require researching the code and regulations with tools, graded by tax professionals.

How it is scored

Accuracy, run by Vals AI.

What to keep in mind

US corporate tax only; rules change yearly.

Top 15
  1. 1Gemini 4 Carbon (high)51.8%
  2. 2Qwen3.8 Max Internal49.7%
  3. 3Claude Fable 5.149.2%
  4. 4Claude Opus 5.545.6%
  5. 5Claude Opus 545.5%
  6. 6Gemini 4 Argon (high)44.6%
  7. 7DeepSeek V4.1 Flash 1017 (max)44.4%
  8. 8Muse Wonder Full (max)43.0%
  9. 9Muse Spark 1.3 (max)42.4%
  10. 10Claude Sonnet 5.542.0%
  11. 11GLM 5.3 (max)39.7%
  12. 12Inkling 2 Small38.9%
  13. 13Claude Fable 538.3%
  14. 14Muse Wonder (max)36.3%
  15. 15Muse Spark 1.3 (xhigh)36.0%
79 of 79
#
1Gemini 4 Carbonhigh51.8%69.2$6.342026-11-05
2Qwen3.8 Max Internal49.7%–$4.072026-11-05
3Claude Fable 5.149.2%66.4$13.182026-11-05
4Claude Opus 5.545.6%63.4$15.092026-11-05
5Claude Opus 545.5%65.2$5.142026-11-05
6Gemini 4 Argonhigh44.6%66.0$3.472026-11-05
7DeepSeek V4.1 Flash 1017max44.4%59.2$0.4592026-11-05
8Muse Wonder Fullmax43.0%62.9$2.152026-11-05
9Muse Spark 1.3max42.4%62.0$0.3902026-11-05
10Claude Sonnet 5.542.0%–$9.252026-11-05
11GLM 5.3max39.7%59.0$1.802026-11-05
12NewInkling 2 Small38.9%57.4$0.1442026-11-05
13Claude Fable 538.3%66.5$6.702026-11-05
14Muse Wondermax36.3%62.3$1.882026-11-05
15Muse Spark 1.3xhigh36.0%58.5$0.3202026-11-05
16Hy434.4%55.7$1.692026-11-05
17Grok 4.7xhigh34.2%58.3$1.802026-11-05
18Kimi K3max33.4%62.1$2.792026-11-05
19Grok 4.6high33.1%60.5$0.9782026-11-05
20Muse Spark 1.2xhigh32.8%61.7$0.2692026-11-05
21MiMo-V2.6-Pro32.4%57.3$0.1342026-11-05
22MiniMax M3.1 Flash32.1%53.0$0.3212026-11-05
23Gemini 3.8 Flashhigh32.1%62.0$0.8572026-11-05
24Qwen3.8 Maxmax32.1%58.8$1.552026-11-05
25Claude Haiku 5.531.9%–$1.162026-11-05
26GLM 5.3 Flashmax31.6%54.2$0.1382026-11-05
27GPT-5.6 Solmax31.1%65.4$7.512026-11-05
28Ember-131.1%–$2.972026-11-05
29Qwen3.8 27Bxhigh30.9%52.5$0.8152026-11-05
30Mistral Large 4 OSShigh30.7%–$0.5032026-11-05
31Mistral Large 4high30.4%49.5$1.392026-11-05
32GPT-5.6 Terramax30.0%60.7$4.832026-11-05
33Claude Opus 4.830.0%60.0$1.692026-11-05
34Claude Sonnet 529.4%55.9$1.792026-11-05
35MiMo-V2.6-Flash28.5%55.0$0.0492026-11-05
36DeepSeek V4 Flashhigh28.1%57.3$0.1562026-11-05
37Step 5 Preview27.3%54.9$0.5532026-11-05
38DeepSeek V4 Promax26.8%57.0$0.7292026-11-05
39DeepSeek V4.1 Flashmax26.6%59.0$0.1352026-11-05
40Muse Spark Openxhigh26.3%54.0$0.8842026-11-05
41Claude Sonnet 4.625.8%58.2$1.192026-11-05
42Kimi K3.1 Promax24.9%–$3.642026-11-05
43Grok 4.5high24.9%56.9$0.4742026-11-05
44GPT-5.6 Lunamax24.8%56.3$0.4652026-11-05
45Claude Opus 4.723.7%63.1$1.272026-11-05
46Gemini 3.7 Flashhigh22.8%62.2$0.4842026-11-05
47GPT-5.5xhigh21.8%64.3$4.072026-11-05
48GPT-6.1 Solmax21.8%67.3$1.322026-11-05
49Gemini 3.5 Flashhigh21.6%62.5$0.8602026-11-05
50GPT-6 Astramax20.7%67.8$5.802026-11-05
51Ling 3.0 Flash Af Rc320.6%–$0.0252026-11-05
52MiniMax M320.6%56.4$0.1642026-11-05
53GPT-6 Lunamax19.9%56.7$0.1912026-11-05
54Mercury 3high18.2%–$0.0462026-11-05
55Gemini 3.6 Flashhigh18.0%61.2$0.3562026-11-05
56GPT-6.1 Terramax17.1%65.3$2.162026-11-05
57Qwen3.7max15.8%60.3$0.4352026-11-05
58Inkling Small15.4%52.1$0.0922026-11-05
59Inkling15.3%47.5$0.3542026-11-05
60GPT-6 Solmax15.0%63.3$2.282026-11-05

More agents & tools benchmarks · All LLM benchmarks

Related benchmarks

Top head-to-heads on Tax Agent Bench

Cite as: BenchLeader, “Tax Agent Bench leaderboard”, data as of 6 Nov 2026, https://www.benchleader.com/benchmarks/vals_tax_agent.

Tax Agent Bench: questions

What does Tax Agent Bench measure?
Corporate tax questions that require researching the code and regulations with tools, graded by tax professionals. Scores are reported in percent of tasks solved; higher is better.
What is the top score on Tax Agent Bench?
As of 6 Nov 2026, the highest score on Tax Agent Bench is 51.8% by Gemini 4 Carbon (result dated 5 Nov 2026), ahead of Qwen3.8 Max Internal at 49.7%. Scores come from Vals AI and are re-read daily.
Which AI model leads Tax Agent Bench?
Gemini 4 Carbon leads Tax Agent Bench with 51.8% as of 6 Nov 2026 (result dated 5 Nov 2026), ahead of Qwen3.8 Max Internal at 49.7%.
How many models have Tax Agent Bench results?
79 model configurations have a Tax Agent Bench result on BenchLeader, all taken from Vals AI.
Who runs Tax Agent Bench and how often is it updated?
Tax Agent Bench is published by Vals AI. BenchLeader re-reads the published results once a day and records the date each result was published.
Does Tax Agent Bench count toward the BenchLeader Index?
No. Tax Agent Bench is shown for reference but left out of the composite index.