Skip to main content

Vals TaxEval v2 (TaxEval v2)

A Vals-created set of questions and responses to tax questions

Data verified

How BenchLM shows TaxEval v2

BenchLM mirrors the public Vals AI TaxEval v2 leaderboard captured from https://www.vals.ai/benchmarks/tax_eval_v2 and updated by Vals on July 9, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.

TaxEval v2 is display only on BenchLM. Vals proprietary or Vals-hosted aggregate views are useful context, but BenchLM does not use them as weighted ranking inputs or as a replacement for benchmark-native source records.

126 Vals rows3 task viewsprivate datasetTasks: Overall, Correctness, Stepwise ReasoningDisplay only

TaxEval v2 score on TaxEval v2 — July 9, 2026

BenchLM mirrors the published taxeval v2 score view for TaxEval v2. Muse Spark 1 1 leads the public snapshot at 79.72% , followed by Muse Spark (77.68%) and Claude Sonnet 4.6 (77.11%). BenchLM does not use these results to rank models overall.

126 modelsExternal benchmark mirrorsCurrentDisplay onlyUpdated July 9, 2026

The published TaxEval v2 snapshot is tightly clustered at the top: Muse Spark 1 1 sits at 79.72%, while the third row is only 2.62 points behind. The broader top-10 spread is 4.09 points, so many of the published scores sit in a relatively narrow band.

126 models have been evaluated on TaxEval v2. The benchmark falls in the External benchmark mirrors category. BenchLM tracks this category separately from its weighted global scoring system, so these results are best compared on the dedicated Korean benchmark views. TaxEval v2 is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About TaxEval v2

Year

2026

Tasks

Tax question answering and response evaluation

Format

Accuracy score

Difficulty

Professional tax reasoning

BenchLM mirrors the public Vals AI TaxEval v2 leaderboard as display-only external evidence. The captured snapshot preserves overall scores, task-level scores where Vals publishes them, uncertainty, latency, and cost-per-test metadata. It is excluded from BenchLM weighted rankings.

BenchLM freshness & provenance

Version

TaxEval v2 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

TaxEval v2 score table (126 models)

1
Muse Spark 1 1meta/muse_spark_1_1
79.72%
2
Muse Sparkmeta/muse_spark
77.68%
3
Claude Sonnet 4.6anthropic/claude-sonnet-4-6
77.11%
4
Claude Fable 5anthropic/claude-fable-5
76.94%
5
GPT-5.6 Terraopenai/gpt-5.6-terra
76.17%
6
Claude Opus 4.6 Thinkinganthropic/claude-opus-4-6-thinking
75.96%
7
Grok 3grok/grok-3
75.88%
8
GPT-5.2openai/gpt-5.2-2025-12-11
75.76%
9
Grok 4 Fast Reasoninggrok/grok-4-fast-reasoning
75.70%
10
Claude Opus 4.8anthropic/claude-opus-4-8
75.63%
11
Claude Sonnet 5anthropic/claude-sonnet-5
75.63%
12
Qwen3.7 Maxalibaba/qwen3.7-max
75.31%
13
Claude Opus 4.7anthropic/claude-opus-4-7
75.27%
14
GPT-5 Miniopenai/gpt-5-mini-2025-08-07
75.22%
15
GPT-4.1openai/gpt-4.1-2025-04-14
75.06%
16
GPT-5.5openai/gpt-5.5
74.98%
17
GPT-5.1openai/gpt-5.1-2025-11-13
74.86%
18
Claude Opus 4.5 20251101 Thinkinganthropic/claude-opus-4-5-20251101-thinking
74.86%
19
O4 Miniopenai/o4-mini-2025-04-16
74.78%
20
GPT-5.6 Solopenai/gpt-5.6-sol
74.78%
21
Qwen3.6 Plusalibaba/qwen3.6-plus
74.73%
22
Kimi K2.6kimi/kimi-k2.6
74.65%
23
O3openai/o3-2025-04-16
74.57%
24
GPT-4oopenai/gpt-4o-2024-11-20
74.53%
25
Gemini 3.5 Flashgoogle/gemini-3.5-flash
74.37%
26
Claude Opus 4.5anthropic/claude-opus-4-5-20251101
74.33%
27
O1openai/o1-2024-12-17
74.28%
28
Kimi K2.5 Thinkingkimi/kimi-k2.5-thinking
74.20%
29
Grok 4.20 0309 Reasoninggrok/grok-4.20-0309-reasoning
74.12%
30
Claude 3.7 Sonnet 20250219 Thinkinganthropic/claude-3-7-sonnet-20250219-thinking
74.04%
31
Qwen3 Max Previewalibaba/qwen3-max-preview
73.96%
32
GPT-5.4openai/gpt-5.4-2026-03-05
73.96%
33
Gemini 3 Flash Previewgoogle/gemini-3-flash-preview
73.88%
34
Mimo V2.5 Proxiaomi/mimo-v2.5-pro
73.79%
35
Claude Opus 4.1 20250805 Thinkinganthropic/claude-opus-4-1-20250805-thinking
73.67%
36
Qwen3 Maxalibaba/qwen3-max
73.51%
37
GPT-5openai/gpt-5-2025-08-07
73.39%
38
GLM 5.2zai/glm-5.2
73.34%
39
Claude Sonnet 4.5 20250929 Thinkinganthropic/claude-sonnet-4-5-20250929-thinking
73.30%
40
Grok 4.1 Fast Reasoninggrok/grok-4-1-fast-reasoning
73.14%
41
Nemotron 3 Ultra 550b A55bnvidia/nemotron-3-ultra-550b-a55b
73.10%
42
Mistral Large 2512mistralai/mistral-large-2512
73.06%
43
Grok 3 Mini Fast Low Reasoninggrok/grok-3-mini-fast-low-reasoning
72.98%
44
Gemini 2.5 Pro Exp 03 25google/gemini-2.5-pro-exp-03-25
72.89%
45
Gemini 3.1 Pro Previewgoogle/gemini-3.1-pro-preview
72.88%
46
Gemini 2.5 Flash Preview 09 2025google/gemini-2.5-flash-preview-09-2025
72.73%
47
MiniMax M3minimax/MiniMax-M3
72.73%
48
Gemini 3 Pro Previewgoogle/gemini-3-pro-preview
72.57%
49
Claude 3.7 Sonnetanthropic/claude-3-7-sonnet-20250219
72.40%
50
Gemini 2.5 Flash Preview 09 2025 Thinkinggoogle/gemini-2.5-flash-preview-09-2025-thinking
72.40%
51
GLM 4.5zai/glm-4.5
72.40%
52
DeepSeek R1fireworks/deepseek-r1
72.28%
53
Qwen3.5 Flashalibaba/qwen3.5-flash
72.16%
54
DeepSeek V4 Prodeepseek/deepseek-v4-pro
72.08%
55
Claude Sonnet 4 20250514 Thinkinganthropic/claude-sonnet-4-20250514-thinking
72.00%
56
GPT-4.1 Miniopenai/gpt-4.1-mini-2025-04-14
71.91%
57
Claude Opus 4anthropic/claude-opus-4-20250514
71.91%
58
Mimo V2.5xiaomi/mimo-v2.5
71.83%
59
Gemini 3.1 Flash Lite Previewgoogle/gemini-3.1-flash-lite-preview
71.79%
60
Kimi K2 Thinkingkimi/kimi-k2-thinking
71.71%
61
GPT Oss 120bfireworks/gpt-oss-120b
71.59%
62
Grok 4 Fast Non Reasoninggrok/grok-4-fast-non-reasoning
71.58%
63
Claude Opus 4.1anthropic/claude-opus-4-1-20250805
71.46%
64
Qwen3.6 27balibaba/qwen3.6-27b
71.26%
65
GPT-5.4 Miniopenai/gpt-5.4-mini-2026-03-17
71.22%
66
GLM 5.1zai/glm-5.1
71.19%
67
Gemini 2.5 Flash Preview 04 17google/gemini-2.5-flash-preview-04-17
71.18%
68
GPT-4oopenai/gpt-4o-2024-08-06
71.14%
69
Grok 3 Mini Fast High Reasoninggrok/grok-3-mini-fast-high-reasoning
71.14%
70
DeepSeek V3 0324fireworks/deepseek-v3-0324
71.10%
71
Grok 4.3grok/grok-4.3
70.81%
72
Qwen3 235b A22bfireworks/qwen3-235b-a22b
70.65%
73
Gemini 2.5 Flash Preview 04 17 Thinkinggoogle/gemini-2.5-flash-preview-04-17-thinking
70.52%
74
Mistral Medium 2505mistralai/mistral-medium-2505
70.32%
75
Moonshotai Kimi K2 Instructtogether/moonshotai/Kimi-K2-Instruct
70.20%
76
Claude 3.5 Sonnetanthropic/claude-3-5-sonnet-20241022
70.16%
77
GLM 5 Thinkingzai/glm-5-thinking
70.03%
78
Gemini 2.0 Flash Thinking Exp 01 21google/gemini-2.0-flash-thinking-exp-01-21
69.79%
79
Claude Sonnet 4anthropic/claude-sonnet-4-20250514
69.62%
80
O3 Miniopenai/o3-mini-2025-01-31
69.42%
81
GLM 4.7zai/glm-4.7
68.77%
82
Gemini 2.0 Pro Exp 02 05google/gemini-2.0-pro-exp-02-05
68.15%
83
MiniMax M2.5minimax/MiniMax-M2.5
68.15%
84
DeepSeek V3p2 Thinkingfireworks/deepseek-v3p2-thinking
68.15%
85
Mistral Medium 3.5mistralai/mistral-medium-3.5
67.99%
86
DeepSeek V3fireworks/deepseek-v3
67.91%
87
Gemini 2.0 Flash Expgoogle/gemini-2.0-flash-exp
67.74%
88
Claude Haiku 4.5 20251001 Thinkinganthropic/claude-haiku-4-5-20251001-thinking
67.54%
89
GPT-5.4 Nanoopenai/gpt-5.4-nano-2026-03-17
67.42%
90
GPT-5 Nanoopenai/gpt-5-nano-2025-08-07
67.38%
91
Grok 2 1212grok/grok-2-1212
67.05%
92
Llama4 Maverick Instruct Basicfireworks/llama4-maverick-instruct-basic
66.56%
93
MiniMax M2.7minimax/MiniMax-M2.7
66.56%
94
MiniMax M2.1minimax/MiniMax-M2.1
66.35%
95
GLM 4.6zai/glm-4.6
66.23%
96
Gemini 2.5 Flash Lite Preview 09 2025google/gemini-2.5-flash-lite-preview-09-2025
66.23%
97
Gemini 2.0 Flash 001google/gemini-2.0-flash-001
65.25%
98
Grok 4 0709grok/grok-4-0709
65.09%
99
Gemini 2.5 Flash Lite Preview 09 2025 Thinkinggoogle/gemini-2.5-flash-lite-preview-09-2025-thinking
64.72%
100
Mistral Large 2411mistralai/mistral-large-2411
63.78%
101
Langston Nim Nvidia Llama 3.3 Nemotron Super 49b V1 42e84561 Thinkingtogether/langston/nim/nvidia/llama-3.3-nemotron-super-49b-v1-42e84561-thinking
63.74%
102
GPT Oss 20bfireworks/gpt-oss-20b
63.70%
103
Magistral Medium 2509mistralai/magistral-medium-2509
61.94%
104
Grok 4.1 Fast Non Reasoninggrok/grok-4-1-fast-non-reasoning
61.86%
105
Command A 03 2025cohere/command-a-03-2025
61.37%
106
Jamba Large 1.6ai21labs/jamba-large-1.6
60.88%
107
Meta Llama Meta Llama 3.1 405B Instruct Turbotogether/meta-llama/Meta-Llama-3.1-405B-Instruct-Turbo
60.88%
108
GPT-4.1 Nanoopenai/gpt-4.1-nano-2025-04-14
60.75%
109
GPT-4o Miniopenai/gpt-4o-mini-2024-07-18
60.55%
110
Magistral Small 2509mistralai/magistral-small-2509
60.30%
111
Langston Nim Nvidia Llama 3.3 Nemotron Super 49b V1 42e84561together/langston/nim/nvidia/llama-3.3-nemotron-super-49b-v1-42e84561
60.22%
112
Laguna Xs.2poolside/laguna-xs.2
59.98%
113
Gemini 1.5 Pro 002google/gemini-1.5-pro-002
59.48%
114
Meta Llama Llama 3.3 70B Instruct Turbotogether/meta-llama/Llama-3.3-70B-Instruct-Turbo
59.44%
115
Mistral Small 2503mistralai/mistral-small-2503
58.30%
116
Jamba 1.5 Largeai21labs/jamba-1.5-large
58.18%
117
Claude 3.5 Haikuanthropic/claude-3-5-haiku-20241022
57.36%
118
Meta Llama Meta Llama 3.1 70B Instruct Turbotogether/meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo
56.17%
119
Meta Llama Llama 4 Scout 17B 16E Instructtogether/meta-llama/Llama-4-Scout-17B-16E-Instruct
55.19%
120
Command A Plus 05 2026cohere/command-a-plus-05-2026
52.82%
121
Mistral Small 2402mistralai/mistral-small-2402
49.14%
122
Gemini 1.5 Flash 002google/gemini-1.5-flash-002
48.20%
123
Jamba Mini 1.6ai21labs/jamba-mini-1.6
44.60%
124
Jamba 1.5 Miniai21labs/jamba-1.5-mini
41.86%
125
Meta Llama Meta Llama 3.1 8B Instruct Turbotogether/meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo
32.34%
126
Laguna M.1poolside/laguna-m.1
1.64%

FAQ

What does TaxEval v2 measure?

A Vals-created set of questions and responses to tax questions

Which model leads the published TaxEval v2 snapshot?

Muse Spark 1 1 currently leads the published TaxEval v2 snapshot with 79.72% taxeval v2 score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on TaxEval v2?

126 AI models are included in BenchLM's mirrored TaxEval v2 snapshot, based on the public leaderboard captured on July 9, 2026.

Last updated: July 9, 2026 · mirrored from the public benchmark leaderboard

The AI models change fast. We track them for you.

For engineers, researchers, and the plain curious — a weekly brief on new models, ranking shifts, and pricing changes.

Free. No spam. Unsubscribe anytime.