Tax Agent Bench leaderboard

Agentic research into US corporate tax questions. Claude Fable 5.1 leads with 49.19%.

Models
73
Leader
Claude Fable 5.1
Source
Vals AI
As of
Oct 24, 2026
  1. 1Claude Fable 5.149.2%
  2. 2Claude Opus 5.545.6%
  3. 3Claude Opus 545.5%
  4. 4Gemini 4 Argon44.6%
  5. 5Muse Wonder Full43.0%
  6. 6Muse Spark 1.3 Max42.4%
  7. 7Claude Sonnet 5.542.0%
  8. 8GLM-5.339.7%
  9. 9Claude Fable 538.3%
  10. 10Muse Wonder36.3%
Tax Agent Bench, accuracy against cost per test
Open weightsProprietaryBest for the price
0%15%30%45%60%$.01$.1$1$10Cost per test, log scale →Claude Fable 5.1Claude Opus 5Muse Spark 1.3 MaxMuse Spark 1.3MiMo-V2.6-ProMiMo-V2.6-FlashLing 3.0 Flash Fin

Full leaderboard

RankModelLabAccuracyCost / testHarnessPrice in / out
2Claude Fable 5.1Anthropic49.19%$13.18—$10 / $50
3Claude Opus 5.5Anthropic45.57%$15.09—$4 / $20
4Claude Opus 5Anthropic45.55%$5.14—$5 / $25
5Gemini 4 ArgonGoogle44.56%$3.47—$2 / $10
7Muse Wonder FullMeta43.00%$2.15——
8Muse Spark 1.3 MaxMeta42.44%$0.39——
9Claude Sonnet 5.5Anthropic41.96%$9.25—$2 / $10
10GLM-5.3Z.ai39.70%$1.80—$0.079 / $1.40
11Claude Fable 5Anthropic38.28%$6.70—$10 / $50
12Muse WonderMeta36.27%$1.88—$4 / $20
13Muse Spark 1.3Meta35.97%$0.32—$1.25 / $4.25
14Grok 4.7xAI34.20%$1.80—$2 / $6
15Kimi K3Moonshot AI33.37%$2.79—$0.24 / $11
16Grok 4.6xAI33.14%$0.98—$2 / $6
17Muse Spark 1.2Meta32.80%$0.27—$1.25 / $4.25
18Hy4 previewTencent32.62%$0.59—$0.75 / $2.25
19MiMo-V2.6-ProXiaomi32.39%$0.13—$0.43 / $0.87
20MiniMax-M3.1-FlashMiniMax32.14%$0.32—$0.15 / $0.60
21Gemini 3.8 FlashGoogle32.12%$0.86—$0.82 / $4.13
22Qwen3.8-MaxAlibaba32.09%$1.55—$2 / $6
23Claude Haiku 5.5Anthropic31.92%$1.16—$0.10 / $0.50
24GLM-5.3-FlashZ.ai31.55%$0.14—$0.15 / $0.50
25GPT-5.6 SolOpenAI31.09%$7.51—$4 / $20
26Ember-1Fireworks AI31.08%$2.97—$3 / $15
27Qwen3.8-27BAlibaba30.89%$0.82—$0.20 / $2.50
28Mistral Large 4Mistral AI30.39%$2.72—$1.36 / $4.18
29GPT-5.6 TerraOpenAI30.04%$4.83—$2 / $12
30Claude Opus 4.8Anthropic30.00%$1.69—$5 / $25
31Claude Sonnet 5Anthropic29.37%$1.79—$2 / $10
32MiMo-V2.6-FlashXiaomi28.51%$0.05—$0.14 / $0.28
33DeepSeek-V4-FlashDeepSeek28.07%$0.16—$0.042 / $0.18
34Step 5 PreviewStepFun27.27%$0.55—$1 / $2.70
36DeepSeek-V4.1-Flash-20260910DeepSeek26.63%$0.13—$0.15 / $0.60
37Muse Spark OpenMeta26.32%$0.88—$1.25 / $4.25
38Claude Sonnet 4.6Anthropic25.84%$1.19—$3 / $15
39DeepSeek-V4-ProDeepSeek25.17%$0.72—$0.66 / $1.98
40Kimi K3.1 ProMoonshot AI24.87%$3.64—$2 / $10
41Grok 4.5xAI24.86%$0.47—$2 / $6
42GPT-5.6 LunaOpenAI24.82%$0.47—$0.20 / $1.20
43Claude Opus 4.7Anthropic23.74%$1.27—$5 / $25
44Gemini 3.7 FlashGoogle22.80%$0.48—$0.75 / $3.75
45GPT-5.5OpenAI21.76%$4.07—$5 / $30
46GPT-6.1 SolOpenAI21.75%$1.32—$2 / $10
47Gemini 3.5 FlashGoogle21.65%$0.86—$1.50 / $9
48GPT-6 AstraOpenAI20.72%$5.80—$10 / $50
49Ling 3.0 Flash Fin—20.60%$0.03——
50MiniMax-M3MiniMax20.59%$0.16—$0.30 / $1.20
51GPT-6 LunaOpenAI19.95%$0.19—$0.10 / $0.50
52Mercury 3Inception18.17%$0.05—$0.10 / $0.35
53Gemini 3.6 FlashGoogle18.04%$0.36—$0.75 / $3.75
54Qwen3.7-Max-PreviewAlibaba15.83%$0.44—$1.48 / $4.42
55Inkling-SmallThinky15.39%$0.09—$0.45 / $1.20
56InklingThinky15.29%$0.35—$1 / $4.05
57GPT-6 SolOpenAI15.00%$2.28—$2 / $10
58Kimi K2.6Moonshot AI12.40%$0.57—$0.65 / $3.41
59Gemini 3 FlashGoogle11.84%$0.10—$0.50 / $3
60MiMo-V2.5-ProXiaomi11.50%$0.04—$0.30 / $0.61
61Qwen3.7-PlusAlibaba10.47%$0.09—$0.32 / $1.28
62Gemini 3.1 Pro PreviewGoogle9.29%$0.30—$2 / $12
63Beam 1.0Reflection AI8.81%N/A——
64GPT-5.4 miniOpenAI8.81%$0.81—$0.75 / $4.50
65MiMo-V2.5Xiaomi7.77%$0.02—$0.12 / $0.24
66Qwen3.6-PlusAlibaba7.68%$0.30—$0.33 / $1.95
67Gemini 3.5 Flash-LiteGoogle7.25%$0.08—$0.30 / $2.50
68Grok 4.20xAI6.95%$0.12—$1.25 / $2.50
69MiniMax-M2.7MiniMax6.87%$0.12—$0.30 / $1.20
70Grok 4.3xAI6.18%$0.12—$1.25 / $2.50
71Claude Haiku 4.5Anthropic5.86%$0.23—$1 / $5
72GPT-5.4 nanoOpenAI4.97%$0.07—$0.20 / $1.25
73Mistral Medium 3.5Mistral AI4.48%$1.44—$1.50 / $7.50
74Gemini 3.1 Flash-Lite PreviewGoogle3.11%$0.01—$0.25 / $1.50
75Nemotron 3.5 LightningFireworks AI2.59%$0.03—$0.07 / $0.20
76Mercury 2.5Inception2.49%$0.02—$0.04 / $0.15

Related leaderboards

Sources

  • Vals AI, as of Oct 24, 2026
  • Provider list prices, as of Oct 24, 2026

Every number on this page comes from these sources; models without a published number show none. Download the data. Cite as “Nexus Explorer, nexus.inc/models”.