Benchmark Almanac

EDINET-Bench

Random Forest 0.75 tops its leaderboard Claude 3.5 Sonnet (October 2024) 0.73 leads the models we track
Not in model cards

Models read Japanese financial disclosures to detect fraud, forecast earnings and identify industries.

The evaluation reports results separately for its three tasks.

Made by
Sakana AI
Released
June 2025
Kind
Finance
Level
Expert
Scored
Checked automatically
Answers
Prediction
Tasks
All public
Measures
Capability
Language
Japanese
In and out
Text, Documents → Text
Interaction
Single turn
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
In views
Law and finance
Strongest model tested
GPT-5 #71 on the capability index
Last model added
September 2025
Scores refreshed
November 20th, 2026
14
Entries in its paper, 8 models we track

Random Forest tops the leaderboard with 0.75 Claude 3.5 Sonnet (October 2024) leads with 0.73 among models released by September 2025

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. The paper gives no overall score, so the headline is its first metric, accounting fraud detection ROC-AUC. 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. The best model we track is Claude 3.5 Sonnet (October 2024), at 0.73. This view shows only models we track, each at its best.

The newest model with a score here came out in September 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Random Forest Summary 0.75
  2. 2 Claude 3.5 Sonnet (October 2024) Summary+BS+CF+PL+Text 0.73
  3. 3 GPT-4o Nov 2024, Summary+BS+CF+PL+Text 0.69
  4. 4 Logistic Summary 0.68
  5. 5 XGBoost Summary 0.68
  6. 6 Claude 3.5 Haiku Summary+BS+CF+PL+Text 0.67
  7. 7 Claude 3.7 Sonnet Summary+BS+CF+PL+Text 0.67
  8. 8 GPT-5 Summary+BS+CF+PL+Text 0.67
  9. 9 Llama-3.2-1B SFT Summary+BS+CF+PL 0.66
  10. 10 DeepSeek-R1 (snapshot unspecified) Summary+BS+CF+PL+Text 0.63
  11. 11 DeepSeek-V3 (snapshot unspecified) Summary 0.61
  12. 12 o4-mini Summary+BS+CF+PL+Text 0.61
  13. 13 Llama 3.3 70B Summary+BS+CF+PL 0.59
  14. 14 Kimi K2 Sep 2025, Summary+BS+CF+PL+Text 0.59

Source: The EDINET-Bench paper

#EntryScoreSettingDateSource
1 Random Forest 0.75 Summary Official leaderboard
2 Claude 3.5 Sonnet (October 2024) 0.73 Summary+BS+CF+PL+Text Oct 2024 Official leaderboard
3 GPT-4o Nov 2024 0.69 Summary+BS+CF+PL+Text Nov 2024 Official leaderboard
4 Logistic 0.68 Summary Official leaderboard
5 XGBoost 0.68 Summary Official leaderboard
6 Claude 3.5 Haiku 0.67 Summary+BS+CF+PL+Text Oct 2024 Official leaderboard
7 Claude 3.7 Sonnet 0.67 Summary+BS+CF+PL+Text Feb 2025 Official leaderboard
8 GPT-5 0.67 Summary+BS+CF+PL+Text Aug 2025 Official leaderboard
9 Llama-3.2-1B SFT 0.66 Summary+BS+CF+PL Official leaderboard
10 DeepSeek-R1 (snapshot unspecified) 0.63 Summary+BS+CF+PL+Text Official leaderboard
11 DeepSeek-V3 (snapshot unspecified) 0.61 Summary Official leaderboard
12 o4-mini 0.61 Summary+BS+CF+PL+Text Apr 2025 Official leaderboard
13 Llama 3.3 70B 0.59 Summary+BS+CF+PL Dec 2024 Official leaderboard
14 Kimi K2 Sep 2025 0.59 Summary+BS+CF+PL+Text Sep 2025 Official leaderboard
#ModelScoreSettingDateSource
1 Claude 3.5 Sonnet (October 2024) 0.73 Summary+BS+CF+PL+Text Oct 2024 Official leaderboard
2 GPT-4o Nov 2024 0.69 Summary+BS+CF+PL+Text Nov 2024 Official leaderboard
3 Claude 3.5 Haiku 0.67 Summary+BS+CF+PL+Text Oct 2024 Official leaderboard
4 Claude 3.7 Sonnet 0.67 Summary+BS+CF+PL+Text Feb 2025 Official leaderboard
5 GPT-5 0.67 Summary+BS+CF+PL+Text Aug 2025 Official leaderboard
6 o4-mini 0.61 Summary+BS+CF+PL+Text Apr 2025 Official leaderboard
7 Llama 3.3 70B 0.59 Summary+BS+CF+PL Dec 2024 Official leaderboard
8 Kimi K2 Sep 2025 0.59 Summary+BS+CF+PL+Text Sep 2025 Official leaderboard

Rankings by part

We read 5 parts from the benchmark’s paper, each ranking models on one part of the test, such as Fraud Detection ROC-AUC, Fraud Detection MCC and Earnings Forecasting ROC-AUC. Claude 3.5 Sonnet (October 2024) leads 3 of them, and GPT-5 leads the other 2. Arrows show how far a model moved from its overall rank among the models we track.

Fraud Detection ROC-AUC

8 models we track have a score here.

  1. 1 Claude 3.5 Sonnet (October 2024) 0.73
  2. 2 GPT-4o Nov 2024 0.69
  3. 3 Claude 3.5 Haiku 0.67
  4. 4 Claude 3.7 Sonnet 0.67
  5. 5 GPT-5 0.67
  6. 6 o4-mini 0.61
  7. 7 Llama 3.3 70B 0.59
  8. 8 Kimi K2 Sep 2025 0.59

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Claude 3.5 Sonnet (October 2024)
Summary+BS+… Summary+BS+…
0.73
GPT-4o
Summary Summary+BS+… Summary+BS+…
0.69
Claude 3.5 Haiku
Summary+BS+… Summary+BS+…
0.67
Claude 3.7 Sonnet
Summary+BS+… Summary+BS+…
0.67
GPT-5
Summary Summary+BS+… Summary+BS+…
0.67
o4-mini
Summary+BS+… Summary+BS+…
0.61
Llama 3.3 70B
Summary+BS+… Summary
0.59
Kimi K2
Summary+BS+… Summary
0.59

Source: The EDINET-Bench paper

ModelEffortHarnessScoreCost per taskOutput tokens
Claude 3.5 Sonnet (October 2024) Summary+BS+CF+PL+Text 0.73
GPT-4o Nov 2024 Summary+BS+CF+PL+Text 0.69
Claude 3.5 Haiku Summary+BS+CF+PL+Text 0.67
Claude 3.7 Sonnet Summary+BS+CF+PL+Text 0.67
GPT-5 Summary+BS+CF+PL+Text 0.67
Claude 3.5 Sonnet (October 2024) Summary 0.64
Claude 3.5 Sonnet (October 2024) Summary+BS+CF+PL 0.63
GPT-5 Summary+BS+CF+PL 0.62
Claude 3.5 Haiku Summary 0.61
GPT-4o Nov 2024 Summary+BS+CF+PL 0.61
o4-mini Summary+BS+CF+PL+Text 0.61
Claude 3.5 Haiku Summary+BS+CF+PL 0.60
GPT-4o Nov 2024 Summary 0.59
Llama 3.3 70B Summary+BS+CF+PL 0.59
Claude 3.7 Sonnet Summary 0.59
Kimi K2 Sep 2025 Summary+BS+CF+PL+Text 0.59
Llama 3.3 70B Summary 0.58
Claude 3.7 Sonnet Summary+BS+CF+PL 0.58
Kimi K2 Sep 2025 Summary 0.58
Llama 3.3 70B Summary+BS+CF+PL+Text 0.56
GPT-5 Summary 0.56
Kimi K2 Sep 2025 Summary+BS+CF+PL 0.54
o4-mini Summary 0.53
o4-mini Summary+BS+CF+PL 0.52

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Claude 3.5 Sonnet (October 2024)
Summary+BS+… Summary+BS+…
0.73
GPT-4o
Summary Summary+BS+… Summary+BS+…
0.69
Claude 3.5 Haiku
Summary+BS+… Summary+BS+…
0.67
Claude 3.7 Sonnet
Summary+BS+… Summary+BS+…
0.67
GPT-5
Summary Summary+BS+… Summary+BS+…
0.67
o4-mini
Summary+BS+… Summary+BS+…
0.61
Llama 3.3 70B
Summary+BS+… Summary
0.59
Kimi K2
Summary+BS+… Summary
0.59

Source: The EDINET-Bench paper

ModelEffortHarnessScoreCost per taskOutput tokens
Claude 3.5 Sonnet (October 2024) Summary+BS+CF+PL+Text 0.73
GPT-4o Nov 2024 Summary+BS+CF+PL+Text 0.69
Claude 3.5 Haiku Summary+BS+CF+PL+Text 0.67
Claude 3.7 Sonnet Summary+BS+CF+PL+Text 0.67
GPT-5 Summary+BS+CF+PL+Text 0.67
Claude 3.5 Sonnet (October 2024) Summary 0.64
Claude 3.5 Sonnet (October 2024) Summary+BS+CF+PL 0.63
GPT-5 Summary+BS+CF+PL 0.62
Claude 3.5 Haiku Summary 0.61
GPT-4o Nov 2024 Summary+BS+CF+PL 0.61
o4-mini Summary+BS+CF+PL+Text 0.61
Claude 3.5 Haiku Summary+BS+CF+PL 0.60
GPT-4o Nov 2024 Summary 0.59
Llama 3.3 70B Summary+BS+CF+PL 0.59
Claude 3.7 Sonnet Summary 0.59
Kimi K2 Sep 2025 Summary+BS+CF+PL+Text 0.59
Llama 3.3 70B Summary 0.58
Claude 3.7 Sonnet Summary+BS+CF+PL 0.58
Kimi K2 Sep 2025 Summary 0.58
Llama 3.3 70B Summary+BS+CF+PL+Text 0.56
GPT-5 Summary 0.56
Kimi K2 Sep 2025 Summary+BS+CF+PL 0.54
o4-mini Summary 0.53
o4-mini Summary+BS+CF+PL 0.52

Number one over time

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

Anthropic
Benchmark paper Not tested since Sep 2025

Source: The EDINET-Bench paper

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
In the benchmark’s paper
0.55 0.60 0.65 0.70 0.75 2025 0.55 0.60 0.65 0.70 0.75 2025

Source: Official EDINET-Bench 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

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 14 months ago, in September 2025, for Kimi K2. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the benchmark’s paper (14 entries).
  • Pass Models scored 8 models we track, plus 6 other entries on its leaderboard.
  • Warning Newest score The newest came 14 months ago, in September 2025, for Kimi K2.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5 (#71).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on sakana.ai: 24 rows for models we track and 6 other entries 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