Benchmark Almanac

DualEntry Accounting AI Benchmark

Gemini 4 Argon 91.1% leads since October 2026
Not in model cards

Models use accounting tools to look up accounts, reconcile balances, and draft entries for a test company. DualEntry checks their answers against fixed references.

DualEntry lists scores for 29 model versions. Each model runs all 101 cases three times. A case counts toward accuracy after two passes.

Made by
DualEntry
Released
October 2026
Kind
Finance
Level
University
Scored
Checked automatically
Answers
Agent task
Measures
Capability
Language
English
In and out
Text → Actions, Text
Interaction
Agentic
Environment
APIs and tools
Refresh
One fixed set
Run by
A host
Occupations
Business and financial operations
Strongest model tested
GPT-6.2 Astra #1 on the capability index
Last model added
November 2026
Scores refreshed
November 20th, 2026
29
Models scored on its own leaderboard

2 models share the top score of 91.1%

Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

  1. 1 Gemini 4 Argon 91.1%
  2. 2 Gemini 4 Neon 91.1%
  3. 3 GLM-5.3-Flash 88.1%
  4. 4 Grok 4.7 88.1%
  5. 5 Gemini 4 Carbon 88.1%
  6. 6 Kimi K3.1 Pro 86.1%
  7. 7 Claude Fable 5.1 85.1%
  8. 8 Kimi K3 84.2%
  9. 9 DeepSeek V4.1 Flash 84.2%
  10. 10 GPT-6.2 Sol 84.2%
  11. 11 Grok 4.5 83.2%
  12. 12 Gemini 3.8 Flash 83.2%
  13. 13 Muse Wonder 83.2%
  14. 14 Gemini 4 Hydrogen 83.2%
  15. 15 Qwen4-Max 83.2%
  16. 16 GLM-5.3 82.2%
  17. 17 Claude Opus 5.5 82.2%
  18. 18 DeepSeek V4.1 Flash 1017 82.2%
  19. 19 GPT-6 Astra 81.2%
  20. 20 GPT-6.1 Sol 81.2%
  21. 21 Claude Fable 5.2 81.2%
  22. 22 Claude Sonnet 5.5 80.2%
  23. 23 Mistral Large 4 80.2%
  24. 24 Mercury 3 78.2%
  25. 25 GPT-6.1 Terra 78.2%
  26. 26 GPT-6.2 Astra 78.2%
  27. 27 GPT-6 Sol 76.2%
  28. 28 Claude Haiku 5.5 76.2%
  29. 29 GPT-6 Luna 66.3%

Source: Official DualEntry Accounting AI Benchmark leaderboard

#ModelScoreSettingDateSource
1 Gemini 4 Argon 91.1% Sep 2026 Official leaderboard
2 Gemini 4 Neon 91.1% Nov 2026 Official leaderboard
3 GLM-5.3-Flash 88.1% Aug 2026 Official leaderboard
4 Grok 4.7 88.1% Sep 2026 Official leaderboard
5 Gemini 4 Carbon 88.1% Oct 2026 Official leaderboard
6 Kimi K3.1 Pro 86.1% Oct 2026 Official leaderboard
7 Claude Fable 5.1 85.1% Sep 2026 Official leaderboard
8 Kimi K3 84.2% Jul 2026 Official leaderboard
9 DeepSeek V4.1 Flash 84.2% Sep 2026 Official leaderboard
10 GPT-6.2 Sol 84.2% Nov 2026 Official leaderboard
11 Grok 4.5 83.2% Jul 2026 Official leaderboard
12 Gemini 3.8 Flash 83.2% Sep 2026 Official leaderboard
13 Muse Wonder 83.2% Oct 2026 Official leaderboard
14 Gemini 4 Hydrogen 83.2% Nov 2026 Official leaderboard
15 Qwen4-Max 83.2% Nov 2026 Official leaderboard
16 GLM-5.3 82.2% Aug 2026 Official leaderboard
17 Claude Opus 5.5 82.2% Sep 2026 Official leaderboard
18 DeepSeek V4.1 Flash 1017 82.2% Oct 2026 Official leaderboard
19 GPT-6 Astra 81.2% Sep 2026 Official leaderboard
20 GPT-6.1 Sol 81.2% Sep 2026 Official leaderboard
21 Claude Fable 5.2 81.2% Nov 2026 Official leaderboard
22 Claude Sonnet 5.5 80.2% Sep 2026 Official leaderboard
23 Mistral Large 4 80.2% Oct 2026 Official leaderboard
24 Mercury 3 78.2% Oct 2026 Official leaderboard
25 GPT-6.1 Terra 78.2% Oct 2026 Official leaderboard
26 GPT-6.2 Astra 78.2% Nov 2026 Official leaderboard
27 GPT-6 Sol 76.2% Sep 2026 Official leaderboard
28 Claude Haiku 5.5 76.2% Oct 2026 Official leaderboard
29 GPT-6 Luna 66.3% Sep 2026 Official leaderboard

Rankings by part

We read 10 parts from the benchmark’s own leaderboard in 2 groups: accounting area and difficulty. Each ranks models on one part of the test. Kimi K3 leads 2 of them, and 7 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Transaction Classification

25 models we track have a score here.

  1. 1 Claude Fable 5.1 ↑6 100%
  2. 2 Gemini 3.8 Flash ↑10 100%
  3. 3 GPT-6 Astra ↑16 100%
  4. 4 DeepSeek V4.1 Flash ↑5 100%
  5. 5 Grok 4.7 ↓1 100%
  6. 6 Claude Opus 5.5 ↑11 100%
  7. 7 Claude Sonnet 5.5 ↑15 100%
  8. 8 GPT-6.1 Sol ↑12 100%
  9. 9 Gemini 4 Argon ↓8 100%
  10. 10 Claude Haiku 5.5 ↑18 100%

The best score, 91.1%, costs $0.11 a task

Runs on the cost frontier score higher than any run that's cheaper. GLM-5.3-Flash comes within 3 points of the best for $0.006 a task, 18 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Z.ai DeepSeek Google DeepMind Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.001 $0.01 $0.10 $1.00 $10 Cost per task, log scale Gemini 4 Argon GLM-5.3-Flash 0% 20% 40% 60% 80% 100% $0.001 $0.01 $0.10 $1.00 $10 Cost per task, log scale Gemini 4 Argon DeepSeek V4.1 Flash

Source: Official DualEntry Accounting AI Benchmark leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 4 Argon 91.1% $0.11
Gemini 4 Neon 91.1% $0.17
GLM-5.3-Flash 88.1% $0.006
Grok 4.7 88.1% $0.13
Gemini 4 Carbon 88.1% $0.08
Kimi K3.1 Pro 86.1% $0.08
Claude Fable 5.1 85.1% $1.18
Kimi K3 84.2% $0.08
DeepSeek V4.1 Flash 84.2% $0.003
GPT-6.2 Sol 84.2% $0.09
Grok 4.5 83.2% $0.09
Gemini 3.8 Flash 83.2% $0.06
Muse Wonder 83.2% $0.11
Gemini 4 Hydrogen 83.2% $0.06
Qwen4-Max 83.2%
GLM-5.3 82.2% $0.04
Claude Opus 5.5 82.2% $0.45
DeepSeek V4.1 Flash 1017 82.2%
GPT-6 Astra 81.2% $0.49
GPT-6.1 Sol 81.2% $0.09
Claude Fable 5.2 81.2% $1.13
Claude Sonnet 5.5 80.2% $0.22
Mistral Large 4 80.2% $0.04
Mercury 3 78.2%
GPT-6.1 Terra 78.2%
GPT-6.2 Astra 78.2% $0.38
GPT-6 Sol 76.2% $0.10
Claude Haiku 5.5 76.2% $0.01
GPT-6 Luna 66.3% $0.005

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Google DeepMind
Own leaderboard Passed 85% in Oct 2026, by Gemini 4 Argon

Source: Official DualEntry Accounting AI Benchmark leaderboard

GLM-5.3-Flash on the official leaderboard passed 85% in August 2026

Models released before it came out already scored that high. All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Gemini 4 Argon GLM-5.3-Flash Grok 4.5 0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Gemini 4 Argon Kimi K3

Source: Official DualEntry Accounting AI Benchmark leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Law, finance and medicine

All 637 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%
  • 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. o1-preview 92.7%
  • Harvey LAB-AA v1.1 In use Artificial Analysis's run of Harvey's Legal Agent Benchmark on Harvey's newer private set of 120 tasks across 24 practice areas. A deliverable with a material hallucination scores zero, and the headline is the share of judges who find every rubric criterion passed. GPT-6.2 Astra 15%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark

Catalog record 1 warning 1 pass
  • Warning Task access We don’t know whether the tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (29 entries).
  • Pass Models scored 29 models we track.
  • Pass Newest score The newest is from November 2026, for Gemini 4 Neon.
  • Pass Frontier models 10 of the top 10 have a score, led by GPT-6.2 Astra (#1).
  • Note Headroom Saturated. The best is 91.1%, past 85% since August 2026.
Leaderboard import 3 pass
  • Pass Importer We read the board on dualentry.com: 29 rows for models we track at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check