# Vals TaxEval v2 (TaxEval v2)

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

Canonical page: https://benchlm.ai/benchmarks/taxevalv2

- Category: [Agentic](/agentic)
- Last updated: September 1, 2026

## About TaxEval v2

- Year: 2026
- Tasks: Tax question answering and response evaluation
- Format: Accuracy score
- Difficulty: Professional tax reasoning
- Paper: [TaxEval v2](https://www.vals.ai/benchmarks/tax_eval_v2)

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.

TaxEval v2 is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (145 models)

| Rank | Model | Configuration | Creator | Score |
|------|-------|---------------|---------|-------|
| 1 | [Muse Spark 1.2](/models/muse-spark-1-2) | xhigh reasoning | Meta | 80.38% |
| 2 | [Muse Spark 1.1](/models/muse-spark-1-1) | xhigh reasoning | Meta | 79.72% |
| 3 | [Muse Spark](/models/muse-spark) | — | Meta | 77.68% |
| 4 | [Claude Sonnet 4.6](/models/claude-sonnet-4-6) | — | Anthropic | 77.11% |
| 5 | [Claude Fable 5](/models/claude-fable) | — | Anthropic | 76.94% |
| 6 | [GPT-5.6 Terra](/models/gpt-5-6-terra) | xhigh reasoning | OpenAI | 76.17% |
| 7 | [GPT-5.6 Luna](/models/gpt-5-6-luna) | max reasoning | OpenAI | 76.17% |
| 8 | [Claude Opus 4.6 (Adaptive)](/models/claude-opus-4-6-thinking) | — | Anthropic | 75.96% |
| 9 | [Claude Fable 5.1](/models/claude-fable-5-1) | — | Anthropic | 75.96% |
| 10 | [Grok 3](https://www.vals.ai/models/grok_grok-3) | — | xAI | 75.88% |
| 11 | [GPT-5.2](/models/gpt-5-2) | xhigh reasoning | OpenAI | 75.76% |
| 12 | [Kimi K3](/models/kimi-k3) | — | Moonshot AI | 75.72% |
| 13 | [Grok 4 Fast (Reasoning)](/models/grok-4-fast-reasoning) | — | xAI | 75.70% |
| 14 | [Claude Opus 4.8](/models/claude-opus-4-8) | — | Anthropic | 75.63% |
| 15 | [Claude Sonnet 5](/models/claude-sonnet-5) | — | Anthropic | 75.63% |
| 16 | [GLM-5.3-Flash](/models/glm-5-3-flash) | max reasoning | Z.AI | 75.59% |
| 17 | [Qwen3.8 Max](/models/qwen3-8-max) | — | Alibaba | 75.55% |
| 18 | [Inkling-Small](/models/inkling-small) | 0.99 reasoning | Thinking Machines Lab | 75.51% |
| 19 | [Qwen3.7 Max](/models/qwen3-7-max) | — | Alibaba | 75.31% |
| 20 | [Inkling](/models/inkling) | 0.99 reasoning | Thinking Machines Lab | 75.31% |
| 21 | [Claude Opus 4.7](/models/claude-opus-4-7) | — | Anthropic | 75.27% |
| 22 | [GPT-5 mini](/models/gpt-5-mini) | high reasoning | OpenAI | 75.22% |
| 23 | [Claude Opus 5](/models/claude-opus-5) | — | Anthropic | 75.14% |
| 24 | [GPT-4.1](/models/gpt-4-1) | high reasoning | OpenAI | 75.06% |
| 25 | [GPT-5.5](/models/gpt-5-5) | xhigh reasoning | OpenAI | 74.98% |
| 26 | [Gemini 3.6 Flash](/models/gemini-3-6-flash) | high reasoning | Google | 74.86% |
| 27 | [GPT-5.1](/models/gpt-5-1) | high reasoning | OpenAI | 74.86% |
| 28 | [Claude Opus 4.5 Thinking](/models/claude-opus-4-5-thinking) | — | Anthropic | 74.86% |
| 29 | [O4 Mini](https://www.vals.ai/models/openai_o4-mini-2025-04-16) | high reasoning | OpenAI | 74.78% |
| 30 | [GPT-5.6 Sol](/models/gpt-5-6-sol) | max reasoning | OpenAI | 74.78% |
| 31 | [Gemini 3.7 Flash](/models/gemini-3-7-flash) | high reasoning | Google | 74.73% |
| 32 | [Qwen3.6 Plus](/models/qwen3-6-plus) | — | Alibaba | 74.73% |
| 33 | [Kimi K2.6](/models/kimi-2-6) | — | Moonshot AI | 74.65% |
| 34 | [o3](/models/o3) | high reasoning | OpenAI | 74.57% |
| 35 | [GPT-4o](/models/gpt-4o) | high reasoning | OpenAI | 74.53% |
| 36 | [Gemini 3.8 Flash](/models/gemini-3-8-flash) | high reasoning | Google | 74.45% |
| 37 | [Gemini 3.5 Flash](/models/gemini-3-5-flash) | high reasoning | Google | 74.37% |
| 38 | [Claude Opus 4.5](/models/claude-opus-4-5) | — | Anthropic | 74.33% |
| 39 | [o1](/models/o1) | high reasoning | OpenAI | 74.28% |
| 40 | [Kimi K2.5 Thinking](https://www.vals.ai/models/kimi_kimi-k2.5-thinking) | — | Moonshot AI | 74.20% |
| 41 | [Grok 4.20 0309 Reasoning](https://www.vals.ai/models/grok_grok-4.20-0309-reasoning) | — | xAI | 74.12% |
| 42 | [Claude 3.7 Sonnet 20250219 Thinking](https://www.vals.ai/models/anthropic_claude-3-7-sonnet-20250219-thinking) | — | Anthropic | 74.04% |
| 43 | [Qwen3 Max Preview](https://www.vals.ai/models/alibaba_qwen3-max-preview) | — | Alibaba | 73.96% |
| 44 | [GPT-5.4](/models/gpt-5-4) | xhigh reasoning | OpenAI | 73.96% |
| 45 | [Gemini 3 Flash Preview](https://www.vals.ai/models/google_gemini-3-flash-preview) | high reasoning | Google | 73.88% |
| 46 | [MiMo-V2.5-Pro](/models/mimo-v2-5-pro) | — | Xiaomi | 73.79% |
| 47 | [Claude Opus 4.1 20250805 Thinking](https://www.vals.ai/models/anthropic_claude-opus-4-1-20250805-thinking) | — | Anthropic | 73.67% |
| 48 | [Qwen3 Max](/models/qwen3-max) | — | Alibaba | 73.51% |
| 49 | [GPT-5](https://www.vals.ai/models/openai_gpt-5-2025-08-07) | high reasoning | OpenAI | 73.39% |
| 50 | [GLM-5.2](/models/glm-5-2) | — | Z.AI | 73.34% |
| 51 | [Claude Sonnet 4.5 Thinking](/models/claude-sonnet-4-5-thinking) | — | Anthropic | 73.30% |
| 52 | [Grok 4.1 Fast (Reasoning)](/models/grok-4-1-fast-reasoning) | — | xAI | 73.14% |
| 53 | [Nemotron 3 Ultra 550b A55b](https://www.vals.ai/models/nvidia_nemotron-3-ultra-550b-a55b) | — | Nvidia | 73.10% |
| 54 | [Mistral Large 2512](https://www.vals.ai/models/mistralai_mistral-large-2512) | — | Mistral AI | 73.06% |
| 55 | [DeepSeek V4 Pro 0813](/models/deepseek-v4-pro-0813) | max reasoning | DeepSeek | 73.06% |
| 56 | [Grok 3 Mini Fast Low Reasoning](https://www.vals.ai/models/grok_grok-3-mini-fast-low-reasoning) | low reasoning | xAI | 72.98% |
| 57 | [Gemini 2.5 Pro Exp 03 25](https://www.vals.ai/models/google_gemini-2.5-pro-exp-03-25) | — | Google | 72.89% |
| 58 | [Gemini 3.1 Pro Preview](https://www.vals.ai/models/google_gemini-3.1-pro-preview) | high reasoning | Google | 72.88% |
| 59 | [Gemini 2.5 Flash Preview 09 2025](https://www.vals.ai/models/google_gemini-2.5-flash-preview-09-2025) | — | Google | 72.73% |
| 60 | [MiniMax M3](/models/minimax-m3) | — | MiniMax | 72.73% |
| 61 | [Gemini 3.5 Flash-Lite](/models/gemini-3-5-flash-lite) | high reasoning | Google | 72.61% |
| 62 | [Gemini 3 Pro Preview](https://www.vals.ai/models/google_gemini-3-pro-preview) | high reasoning | Google | 72.57% |
| 63 | [Claude 3.7 Sonnet](https://www.vals.ai/models/anthropic_claude-3-7-sonnet-20250219) | — | Anthropic | 72.40% |
| 64 | [Gemini 2.5 Flash Preview 09 2025 Thinking](https://www.vals.ai/models/google_gemini-2.5-flash-preview-09-2025-thinking) | — | Google | 72.40% |
| 65 | [GLM-4.5](/models/glm-4-5) | — | Z.AI | 72.40% |
| 66 | [GLM-5.3](/models/glm-5-3) | max reasoning | Z.AI | 72.36% |
| 67 | [DeepSeek-R1](/models/deepseek-r1) | — | DeepSeek | 72.28% |
| 68 | [Qwen3.5 Flash](/models/qwen3-5-flash) | — | Alibaba | 72.16% |
| 69 | [DeepSeek V4 Pro 0813](/models/deepseek-v4-pro-0813) | max reasoning | DeepSeek | 72.08% |
| 70 | [Claude Sonnet 4 20250514 Thinking](https://www.vals.ai/models/anthropic_claude-sonnet-4-20250514-thinking) | — | Anthropic | 72.00% |
| 71 | [GPT-4.1 mini](/models/gpt-4-1-mini) | high reasoning | OpenAI | 71.91% |
| 72 | [Claude Opus 4](https://www.vals.ai/models/anthropic_claude-opus-4-20250514) | — | Anthropic | 71.91% |
| 73 | [MiMo-V2.5](/models/mimo-v2-5) | — | Xiaomi | 71.83% |
| 74 | [Gemini 3.1 Flash Lite Preview](https://www.vals.ai/models/google_gemini-3.1-flash-lite-preview) | high reasoning | Google | 71.79% |
| 75 | [Kimi K2 Thinking](https://www.vals.ai/models/kimi_kimi-k2-thinking) | — | Moonshot AI | 71.71% |
| 76 | [Grok 4.5](/models/grok-4-5) | high reasoning | xAI | 71.67% |
| 77 | [GPT-OSS 120B](/models/gpt-oss-120b) | — | OpenAI | 71.59% |
| 78 | [Grok 4 Fast Non Reasoning](https://www.vals.ai/models/grok_grok-4-fast-non-reasoning) | — | xAI | 71.58% |
| 79 | [Claude Opus 4.1](https://www.vals.ai/models/anthropic_claude-opus-4-1-20250805) | — | Anthropic | 71.46% |
| 80 | [Qwen3.6-27B](/models/qwen3-6-27b) | — | Alibaba | 71.26% |
| 81 | [GPT-5.4 mini](/models/gpt-5-4-mini) | xhigh reasoning | OpenAI | 71.22% |
| 82 | [GLM-5.1](/models/glm-5-1) | — | Z.AI | 71.19% |
| 83 | [Gemini 2.5 Flash Preview 04 17](https://www.vals.ai/models/google_gemini-2.5-flash-preview-04-17) | — | Google | 71.18% |
| 84 | [GPT-4o](/models/gpt-4o) | high reasoning | OpenAI | 71.14% |
| 85 | [Grok 3 Mini Fast High Reasoning](https://www.vals.ai/models/grok_grok-3-mini-fast-high-reasoning) | high reasoning | xAI | 71.14% |
| 86 | [DeepSeek V3 0324](https://www.vals.ai/models/fireworks_deepseek-v3-0324) | — | Fireworks AI | 71.10% |
| 87 | [Grok 4.6](/models/grok-4-6) | high reasoning | xAI | 71.10% |
| 88 | [Qwen3.8-27B](/models/qwen3-8-27b) | xhigh reasoning | Alibaba | 70.85% |
| 89 | [Grok 4.3](/models/grok-4-3) | — | xAI | 70.81% |
| 90 | [DeepSeek V4 Flash 0731](/models/deepseek-v4-flash-0731) | high reasoning | DeepSeek | 70.69% |
| 91 | [Ling 3.0 Flash 2607](https://www.vals.ai/models/ant_ling-3.0-flash-2607) | — | Ant | 70.65% |
| 92 | [Qwen3 235b A22b](https://www.vals.ai/models/fireworks_qwen3-235b-a22b) | — | Fireworks AI | 70.65% |
| 93 | [Gemini 2.5 Flash Preview 04 17 Thinking](https://www.vals.ai/models/google_gemini-2.5-flash-preview-04-17-thinking) | — | Google | 70.52% |
| 94 | [Mistral Medium 2505](https://www.vals.ai/models/mistralai_mistral-medium-2505) | — | Mistral AI | 70.32% |
| 95 | [Moonshotai Kimi K2 Instruct](https://www.vals.ai/models/together_moonshotai/Kimi-K2-Instruct) | — | Together AI | 70.20% |
| 96 | [Claude 3.5 Sonnet](/models/claude-3-5-sonnet) | — | Anthropic | 70.16% |
| 97 | [GLM 5 Thinking](https://www.vals.ai/models/zai_glm-5-thinking) | — | Zhipu AI | 70.03% |
| 98 | [Gemini 2.0 Flash Thinking Exp 01 21](https://www.vals.ai/models/google_gemini-2.0-flash-thinking-exp-01-21) | — | Google | 69.79% |
| 99 | [Claude Sonnet 4](https://www.vals.ai/models/anthropic_claude-sonnet-4-20250514) | — | Anthropic | 69.62% |
| 100 | [o3-mini](/models/o3-mini) | high reasoning | OpenAI | 69.42% |
| 101 | [GLM-4.7](/models/glm-4-7) | — | Z.AI | 68.77% |
| 102 | [Gemini 2.0 Pro Exp 02 05](https://www.vals.ai/models/google_gemini-2.0-pro-exp-02-05) | — | Google | 68.15% |
| 103 | [MiniMax M2.5](/models/minimax-m2-5) | — | MiniMax | 68.15% |
| 104 | [DeepSeek V3p2 Thinking](https://www.vals.ai/models/fireworks_deepseek-v3p2-thinking) | high reasoning | Fireworks AI | 68.15% |
| 105 | [Mistral Medium 3.5](https://www.vals.ai/models/mistralai_mistral-medium-3.5) | high reasoning | Mistral AI | 67.99% |
| 106 | [DeepSeek V3](/models/deepseek-v3) | — | DeepSeek | 67.91% |
| 107 | [Gemini 2.0 Flash Exp](https://www.vals.ai/models/google_gemini-2.0-flash-exp) | — | Google | 67.74% |
| 108 | [Claude Haiku 4.5 Thinking](/models/claude-haiku-4-5-thinking) | — | Anthropic | 67.54% |
| 109 | [GPT-5.4 nano](/models/gpt-5-4-nano) | high reasoning | OpenAI | 67.42% |
| 110 | [GPT-5 nano](/models/gpt-5-nano) | high reasoning | OpenAI | 67.38% |
| 111 | [Grok 2 1212](https://www.vals.ai/models/grok_grok-2-1212) | — | xAI | 67.05% |
| 112 | [Llama4 Maverick Instruct Basic](https://www.vals.ai/models/fireworks_llama4-maverick-instruct-basic) | — | Fireworks AI | 66.56% |
| 113 | [MiniMax M2.7](/models/minimax-m2-7) | — | MiniMax | 66.56% |
| 114 | [MiniMax M2.1](https://www.vals.ai/models/minimax_MiniMax-M2.1) | — | MiniMax | 66.35% |
| 115 | [GLM-4.6](/models/glm-4-6) | — | Z.AI | 66.23% |
| 116 | [Gemini 2.5 Flash Lite Preview 09 2025](https://www.vals.ai/models/google_gemini-2.5-flash-lite-preview-09-2025) | — | Google | 66.23% |
| 117 | [Gemini 2.0 Flash 001](https://www.vals.ai/models/google_gemini-2.0-flash-001) | — | Google | 65.25% |
| 118 | [Grok 4 0709](https://www.vals.ai/models/grok_grok-4-0709) | — | xAI | 65.09% |
| 119 | [Gemini 2.5 Flash Lite Preview 09 2025 Thinking](https://www.vals.ai/models/google_gemini-2.5-flash-lite-preview-09-2025-thinking) | — | Google | 64.72% |
| 120 | [Mistral Large 2411](https://www.vals.ai/models/mistralai_mistral-large-2411) | — | Mistral AI | 63.78% |
| 121 | [Langston Nim Nvidia Llama 3.3 Nemotron Super 49b V1 42e84561 Thinking](https://www.vals.ai/models/together_langston/nim/nvidia/llama-3.3-nemotron-super-49b-v1-42e84561-thinking) | — | Together AI | 63.74% |
| 122 | [GPT-OSS 20B](/models/gpt-oss-20b) | — | OpenAI | 63.70% |
| 123 | [Magistral Medium 2509](https://www.vals.ai/models/mistralai_magistral-medium-2509) | — | Mistral AI | 61.94% |
| 124 | [Grok 4.1 Fast Non Reasoning](https://www.vals.ai/models/grok_grok-4-1-fast-non-reasoning) | — | xAI | 61.86% |
| 125 | [Command A 03 2025](https://www.vals.ai/models/cohere_command-a-03-2025) | — | Cohere | 61.37% |
| 126 | [Jamba Large 1.6](https://www.vals.ai/models/ai21labs_jamba-large-1.6) | — | AI21 Labs | 60.88% |
| 127 | [Meta Llama Meta Llama 3.1 405B Instruct Turbo](https://www.vals.ai/models/together_meta-llama/Meta-Llama-3.1-405B-Instruct-Turbo) | — | Together AI | 60.88% |
| 128 | [GPT-4.1 nano](/models/gpt-4-1-nano) | high reasoning | OpenAI | 60.75% |
| 129 | [GPT-4o mini](/models/gpt-4o-mini) | high reasoning | OpenAI | 60.55% |
| 130 | [Magistral Small 2509](https://www.vals.ai/models/mistralai_magistral-small-2509) | — | Mistral AI | 60.30% |
| 131 | [Langston Nim Nvidia Llama 3.3 Nemotron Super 49b V1 42e84561](https://www.vals.ai/models/together_langston/nim/nvidia/llama-3.3-nemotron-super-49b-v1-42e84561) | — | Together AI | 60.22% |
| 132 | [Gemini 1.5 Pro 002](https://www.vals.ai/models/google_gemini-1.5-pro-002) | — | Google | 59.48% |
| 133 | [Meta Llama Llama 3.3 70B Instruct Turbo](https://www.vals.ai/models/together_meta-llama/Llama-3.3-70B-Instruct-Turbo) | — | Together AI | 59.44% |
| 134 | [Laguna XS.2](/models/laguna-xs-2) | — | Poolside | 58.95% |
| 135 | [Mistral Small 2503](https://www.vals.ai/models/mistralai_mistral-small-2503) | — | Mistral AI | 58.30% |
| 136 | [Jamba 1.5 Large](https://www.vals.ai/models/ai21labs_jamba-1.5-large) | — | AI21 Labs | 58.18% |
| 137 | [Claude 3.5 Haiku](https://www.vals.ai/models/anthropic_claude-3-5-haiku-20241022) | — | Anthropic | 57.36% |
| 138 | [Meta Llama Meta Llama 3.1 70B Instruct Turbo](https://www.vals.ai/models/together_meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo) | — | Together AI | 56.17% |
| 139 | [Meta Llama Llama 4 Scout 17B 16E Instruct](https://www.vals.ai/models/together_meta-llama/Llama-4-Scout-17B-16E-Instruct) | — | Together AI | 55.19% |
| 140 | [Mistral Small 2402](https://www.vals.ai/models/mistralai_mistral-small-2402) | — | Mistral AI | 49.14% |
| 141 | [Gemini 1.5 Flash 002](https://www.vals.ai/models/google_gemini-1.5-flash-002) | — | Google | 48.20% |
| 142 | [Jamba Mini 1.6](https://www.vals.ai/models/ai21labs_jamba-mini-1.6) | — | AI21 Labs | 44.60% |
| 143 | [Jamba 1.5 Mini](https://www.vals.ai/models/ai21labs_jamba-1.5-mini) | — | AI21 Labs | 41.86% |
| 144 | [Meta Llama Meta Llama 3.1 8B Instruct Turbo](https://www.vals.ai/models/together_meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo) | — | Together AI | 32.34% |
| 145 | [Laguna M.1](/models/laguna-m-1) | — | Poolside | 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 1.2 currently leads the published TaxEval v2 snapshot with a score of 80.38%.

### How many models are evaluated on TaxEval v2?

The September 1, 2026 contains 145 AI models.

### Does TaxEval v2 affect BenchLM's overall score?

Not directly. TaxEval v2 is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.