TaxCalcBench v2
Muse Wonder 83.7% leads since October 2026 Models read W-2, 1099 and other PDFs and compute full federal and state returns line by line. A return counts only when every evaluated line is exact.
v2 adds realistic PDF inputs, state returns and more complex cases, so scores differ from v1. 50 returns. Fireworks lists 14 models.
- Made by
- Column Tax
- Released
- June 2026
- Kind
- Finance
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- All public
- Strongest model tested
- Claude Fable 5.2 #1 on the capability index
- Last model added
- November 2026
- Scores refreshed
- November 12th, 2026
34
Entries on its own leaderboard, 32 models we track
Muse Wonder leads with 83.7%
Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.
- 83.7 Muse Wonder
- 82.0 Claude Fable 5.2
- 79.6
Gemini 4 Argon - 76.0 GPT-6 Sol
- 74.0 GPT-6.1 Sol
- 72.0 GPT-5.6 Sol
- 71.4 Gemini 4 Carbon
- 70.0 GPT-6 Astra
- 70.0 Claude Opus 5.5
- 68.0 GPT-5.5
- 66.0 Kimi K3.1 Pro High Speed
- 62.0 GPT-6.1 Terra
- 58.0 Claude Fable 5.1
- 54.0 Meta Muse Ember
- 54.0 Claude Opus 5
- 50.0 Claude Sonnet 5.5
- 48.0 Grok 4.7
- 46.3 Gemini 3.8 Flash
- 46.0 Claude Fable 5
- 42.0 Kimi K3
- 38.0 Claude Opus 4.8
- 36.0 GPT-6 Luna
- 32.0 Muse Spark 1.3
- 26.0 Gemini 3.7 Flash
- 24.0 Claude Haiku 5.5
- 22.0 Gemini 3.6 Flash
- 20.0 Muse Spark 1.2
- 6.00 Claude Sonnet 5
- 6.00
Mistral Large 4 - 4.00
DeepSeek V4.1 Flash - 2.00 Gemini 3.1 Pro
- 2.00 Gemini 3.5 Flash
- 0.0 GLM-5.3
- 0.0
MiniMax-M3.1-Flash
- 1 Muse Wonder web search 83.7%
- 2 Claude Fable 5.2 web search 82%
- 3 Gemini 4 Argon web search 79.6%
- 4 GPT-6 Sol web search 76%
- 5 GPT-6.1 Sol web search 74%
- 6 GPT-5.6 Sol web search 72%
- 7 Gemini 4 Carbon web search 71.4%
- 8 GPT-6 Astra web search 70%
- 9 Claude Opus 5.5 web search 70%
- 10 GPT-5.5 web search 68%
- 11 Kimi K3.1 Pro High Speed web search 66%
- 12 GPT-6.1 Terra web search 62%
- 13 Claude Fable 5.1 web search 58%
- 14 Meta Muse Ember web search 54%
- 15 Claude Opus 5 web search 54%
- 16 Claude Sonnet 5.5 web search 50%
- 17 Grok 4.7 web search 48%
- 18 Gemini 3.8 Flash web search 46.3%
- 19 Claude Fable 5 web search 46%
- 20 Kimi K3 web search 42%
- 21 Claude Opus 4.8 web search 38%
- 22 GPT-6 Luna web search 36%
- 23 Muse Spark 1.3 web search 32%
- 24 Gemini 3.7 Flash web search 26%
- 25 Claude Haiku 5.5 web search 24%
- 26 Gemini 3.6 Flash web search 22%
- 27 Muse Spark 1.2 web search 20%
- 28 Claude Sonnet 5 6.00%
- 29 Mistral Large 4 web search 6.00%
- 30 DeepSeek V4.1 Flash 4.00%
- 31 Gemini 3.1 Pro 2.00%
- 32 Gemini 3.5 Flash 2.00%
- 33 GLM-5.3 0%
- 34 MiniMax-M3.1-Flash 0%
Source: Official TaxCalcBench v2 leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Muse Wonder | 83.7% | web search | Oct 2026 | Official leaderboard |
| 2 | Claude Fable 5.2 | 82% | web search | Nov 2026 | Official leaderboard |
| 3 | Gemini 4 Argon | 79.6% | web search | Sep 2026 | Official leaderboard |
| 4 | GPT-6 Sol | 76% | web search | Sep 2026 | Official leaderboard |
| 5 | GPT-6.1 Sol | 74% | web search | Sep 2026 | Official leaderboard |
| 6 | GPT-5.6 Sol | 72% | web search | Jul 2026 | Official leaderboard |
| 7 | Gemini 4 Carbon | 71.4% | web search | Oct 2026 | Official leaderboard |
| 8 | GPT-6 Astra | 70% | web search | Sep 2026 | Official leaderboard |
| 9 | Claude Opus 5.5 | 70% | web search | Sep 2026 | Official leaderboard |
| 10 | GPT-5.5 | 68% | web search | Apr 2026 | Official leaderboard |
| 11 | Kimi K3.1 Pro High Speed | 66% | web search | Official leaderboard | |
| 12 | GPT-6.1 Terra | 62% | web search | Oct 2026 | Official leaderboard |
| 13 | Claude Fable 5.1 | 58% | web search | Sep 2026 | Official leaderboard |
| 14 | Meta Muse Ember | 54% | web search | Official leaderboard | |
| 15 | Claude Opus 5 | 54% | web search | Jul 2026 | Official leaderboard |
| 16 | Claude Sonnet 5.5 | 50% | web search | Sep 2026 | Official leaderboard |
| 17 | Grok 4.7 | 48% | web search | Sep 2026 | Official leaderboard |
| 18 | Gemini 3.8 Flash | 46.3% | web search | Sep 2026 | Official leaderboard |
| 19 | Claude Fable 5 | 46% | web search | Jun 2026 | Official leaderboard |
| 20 | Kimi K3 | 42% | web search | Jul 2026 | Official leaderboard |
| 21 | Claude Opus 4.8 | 38% | web search | May 2026 | Official leaderboard |
| 22 | GPT-6 Luna | 36% | web search | Sep 2026 | Official leaderboard |
| 23 | Muse Spark 1.3 | 32% | web search | Sep 2026 | Official leaderboard |
| 24 | Gemini 3.7 Flash | 26% | web search | Aug 2026 | Official leaderboard |
| 25 | Claude Haiku 5.5 | 24% | web search | Oct 2026 | Official leaderboard |
| 26 | Gemini 3.6 Flash | 22% | web search | Jul 2026 | Official leaderboard |
| 27 | Muse Spark 1.2 | 20% | web search | Aug 2026 | Official leaderboard |
| 28 | Claude Sonnet 5 | 6.00% | Jun 2026 | Official leaderboard | |
| 29 | Mistral Large 4 | 6.00% | web search | Oct 2026 | Official leaderboard |
| 30 | DeepSeek V4.1 Flash | 4.00% | Sep 2026 | Official leaderboard | |
| 31 | Gemini 3.1 Pro | 2.00% | Feb 2026 | Official leaderboard | |
| 32 | Gemini 3.5 Flash | 2.00% | May 2026 | Official leaderboard | |
| 33 | GLM-5.3 | 0% | Aug 2026 | Official leaderboard | |
| 34 | MiniMax-M3.1-Flash | 0% | Oct 2026 | Official leaderboard |
The best score, 83.7%, costs $2.71 a task
Runs on the cost frontier score higher than any run that's cheaper. Gemini 4 Argon at web search comes within 4.1 points of the best for $0.88 a task, 3 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.
Google DeepMind OpenAI Anthropic Meta
Cost frontier
One model at each effort level
Source: Official TaxCalcBench v2 leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Muse Wonder | web search | 83.7% | $2.71 | ||
| Claude Fable 5.2 | web search | 82% | $2.62 | ||
| Gemini 4 Argon | web search | 79.6% | $0.88 | ||
| GPT-6 Sol | web search | 76% | $0.36 | ||
| GPT-6.1 Sol | web search | 74% | $0.32 | ||
| GPT-5.6 Sol | web search | 72% | |||
| Gemini 4 Carbon | web search | 71.4% | $1.05 | ||
| GPT-6 Astra | web search | 70% | $1.59 | ||
| Claude Opus 5.5 | web search | 70% | $3.54 | ||
| GPT-5.5 | web search | 68% | |||
| GPT-6.1 Terra | web search | 62% | $0.57 | ||
| Claude Fable 5.2 | 62% | $1.02 | |||
| Claude Fable 5.1 | web search | 58% | $5.28 | ||
| Gemini 4 Carbon | 54.2% | $0.93 | |||
| Claude Opus 5 | web search | 54% | $2.15 | ||
| Claude Sonnet 5.5 | web search | 50% | $2.39 | ||
| Grok 4.7 | web search | 48% | $2.53 | ||
| Gemini 3.8 Flash | web search | 46.3% | $1.06 | ||
| Claude Fable 5 | web search | 46% | |||
| GPT-6 Astra | 46% | $1.04 | |||
| Claude Fable 5.1 | 44% | $2.36 | |||
| Gemini 4 Argon | 42.9% | $0.27 | |||
| Kimi K3 | web search | 42% | $0.58 | ||
| Claude Opus 5.5 | 42% | $1.12 | |||
| GPT-6.1 Sol | 40% | $0.19 | |||
| Claude Opus 4.8 | web search | 38% | |||
| GPT-6 Luna | web search | 36% | $0.23 | ||
| Claude Fable 5 | 32% | ||||
| GPT-5.6 Sol | 32% | ||||
| Muse Spark 1.3 | web search | 32% | $0.53 | ||
| GPT-6 Sol | 30% | $0.24 | |||
| Gemini 3.7 Flash | web search | 26% | $0.35 | ||
| GPT-5.5 | 24% | ||||
| Claude Haiku 5.5 | web search | 24% | $0.34 | ||
| Muse Wonder | 24% | $0.75 | |||
| GPT-6.1 Terra | 24% | $0.18 | |||
| Gemini 3.6 Flash | web search | 22% | $0.53 | ||
| Claude Opus 4.8 | 20% | ||||
| Claude Opus 5 | 20% | $0.59 | |||
| Muse Spark 1.2 | web search | 20% | $1.07 | ||
| Claude Sonnet 5.5 | 18% | $0.31 | |||
| Gemini 3.7 Flash | 16% | $0.07 | |||
| GPT-6 Luna | 14% | $0.03 | |||
| Gemini 3.6 Flash | 8.00% | ||||
| Grok 4.7 | 8.00% | $0.41 | |||
| Claude Sonnet 5 | 6.00% | ||||
| Kimi K3 | 6.00% | ||||
| Muse Spark 1.2 | 6.00% | $0.07 | |||
| Gemini 3.8 Flash | 6.00% | $0.16 | |||
| Muse Spark 1.3 | 6.00% | $0.07 | |||
| Mistral Large 4 | web search | 6.00% | $0.12 | ||
| DeepSeek V4.1 Flash | 4.00% | $0.06 | |||
| Gemini 3.1 Pro | 2.00% | ||||
| Gemini 3.5 Flash | 2.00% | ||||
| Claude Haiku 5.5 | 2.00% | $0.02 | |||
| GLM-5.3 | 0% | $0.42 | |||
| Mistral Large 4 | 0% | $0.06 | |||
| MiniMax-M3.1-Flash | 0% | $0.03 |
5 models have held number one on the official leaderboard since its release in June 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
OpenAI Google DeepMind Meta
Source: Official TaxCalcBench v2 leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
On the benchmark’s own leaderboard
Best so far
Source: Official TaxCalcBench v2 leaderboard
No lab model card reports it yet
We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.
Versions
We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.
Google DeepMind OpenAI Anthropic Meta
Source: Official leaderboards
More in Law, finance and medicine
- MedXpertQA-MM In use Models answer expert-level medical exam questions that combine images with clinical records and test results. Scoring is accuracy.
Qwen 3.8 Max 80.4% - HealthBench Professional In use 525 physician-authored tasks from real clinician chats (care consult, documentation, research) with physician-adjudicated rubrics. Claude Sonnet 5.5 69.2%
- HealthBench In use 5,000 realistic health conversations graded against physician-written rubrics. A length-adjusted variant penalizes verbosity. GPT-5 67.2%
- FinSearchComp In use Agents search the web to fetch live financial data, look up history, and run multi-step investigations, across 635 expert-written questions. Scored by answer accuracy.
Ling-3.1-flash 81.7% - KBL In use Models answer Korean legal knowledge and reasoning tasks and Korean bar exam questions. It has 7 knowledge tasks, 4 reasoning tasks and 2,510 bar exam items.
Solar Pro 4 77.5% - KorMedMCQA In use Models answer 7,469 multiple-choice questions from Korean licensing exams for doctors, nurses, pharmacists and dentists, held from 2012 to 2024. Solar Pro 4 93.2%