LongJud-Bench
Agents conduct connected stages of Chinese civil litigation in LegalWorld. The authors compare their legal consultations, court documents, and courtroom advocacy.
- Made by
- Songhan Zuo and colleagues
- Released
- June 2026
- Kind
- Law
- Level
- Expert
- Scored
- Graded by a model
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Subject
- Law
- Language
- Chinese
- Interaction
- Multi-agent
- Environment
- LegalWorld, APIs and tools
- Refresh
- One fixed set
- Run by
- Anyone
- Occupations
- Legal
- In views
- Agents, Robotics, Law and finance
- Strongest model tested
- GPT-5.2 #51 on the capability index
- Last model added
- April 2026
- Scores refreshed
- November 20th, 2026
Kimi K2.5 leads with 0.66 among models released by April 2026
Best score per model from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. The cross-model table reports no overall score, so this is the plain mean of eight capabilities, averaging the two instances where applicable. Where a model ran with several settings or harnesses, we keep its best.
The newest model with a score here came out in April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1 Kimi K2.5 0.66
- 2
Qwen3.5-Plus 0.65
- 3 GPT-5.2 0.63
- 4
DeepSeek-V4-Flash 0.63
- 5 GLM-4.7 0.61
- 6
Qwen3.5-Flash 0.55
Source: The LongJud-Bench paper
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Kimi K2.5 | 0.66 | LegalWorld | Jan 2026 | Official leaderboard |
| 2 | | 0.65 | LegalWorld | Feb 2026 | Official leaderboard |
| 3 | GPT-5.2 | 0.63 | LegalWorld | Dec 2025 | Official leaderboard |
| 4 | | 0.63 | LegalWorld | Apr 2026 | Official leaderboard |
| 5 | GLM-4.7 | 0.61 | LegalWorld | Dec 2025 | Official leaderboard |
| 6 | | 0.55 | LegalWorld | Feb 2026 | Official leaderboard |
Rankings by part
We read 22 parts from the benchmark’s paper in 5 groups: legal consultation, document drafting, first instance, second instance and courtroom advocacy. Each ranks models on one part of the test. Kimi K2.5 leads 13 of them, and 3 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Issue Spotting
6 models we track have a score here.
- 1 Kimi K2.5 0.67
- 2 GPT-5.2 ↑1 0.63
- 3
Qwen3.5-Plus ↓1 0.62
- 4
DeepSeek-V4-Flash 0.62
- 5
Qwen3.5-Flash ↑1 0.56
- 6 GLM-4.7 ↓1 0.54
Party Identification
6 models we track have a score here.
- 1 Kimi K2.5 0.82
- 2
Qwen3.5-Plus 0.78
- 3
DeepSeek-V4-Flash ↑1 0.73
- 4 GLM-4.7 ↑1 0.69
- 5 GPT-5.2 ↓2 0.68
- 6
Qwen3.5-Flash 0.64
Claim Construction
6 models we track have a score here.
- 1
Qwen3.5-Plus ↑1 0.72
- 2 Kimi K2.5 ↓1 0.70
- 3 GLM-4.7 ↑2 0.66
- 4 GPT-5.2 ↓1 0.65
- 5
DeepSeek-V4-Flash ↓1 0.65
- 6
Qwen3.5-Flash 0.63
Fact Marshalling
6 models we track have a score here.
- 1 Kimi K2.5 0.71
- 2 GLM-4.7 ↑3 0.69
- 3
Qwen3.5-Plus ↓1 0.68
- 4
DeepSeek-V4-Flash 0.68
- 5
Qwen3.5-Flash ↑1 0.67
- 6 GPT-5.2 ↓3 0.64
Evidence Marshalling
6 models we track have a score here.
- 1 GLM-4.7 ↑4 0.71
- 2
Qwen3.5-Plus 0.69
- 3 GPT-5.2 0.67
- 4 Kimi K2.5 ↓3 0.63
- 5
DeepSeek-V4-Flash ↓1 0.63
- 6
Qwen3.5-Flash 0.52
Party Identification
6 models we track have a score here.
- 1 Kimi K2.5 0.80
- 2
Qwen3.5-Plus 0.71
- 3
DeepSeek-V4-Flash ↑1 0.64
- 4 GLM-4.7 ↑1 0.56
- 5 GPT-5.2 ↓2 0.53
- 6
Qwen3.5-Flash 0.46
Claim Construction
6 models we track have a score here.
- 1
Qwen3.5-Plus ↑1 0.72
- 2 Kimi K2.5 ↓1 0.69
- 3 GLM-4.7 ↑2 0.66
- 4
DeepSeek-V4-Flash 0.65
- 5 GPT-5.2 ↓2 0.60
- 6
Qwen3.5-Flash 0.56
Fact Marshalling
6 models we track have a score here.
- 1 Kimi K2.5 0.68
- 2 GLM-4.7 ↑3 0.67
- 3
DeepSeek-V4-Flash ↑1 0.67
- 4
Qwen3.5-Plus ↓2 0.66
- 5
Qwen3.5-Flash ↑1 0.62
- 6 GPT-5.2 ↓3 0.60
Evidence Marshalling
6 models we track have a score here.
- 1 Kimi K2.5 0.74
- 2 GLM-4.7 ↑3 0.72
- 3 GPT-5.2 0.71
- 4
Qwen3.5-Plus ↓2 0.69
- 5
DeepSeek-V4-Flash ↓1 0.69
- 6
Qwen3.5-Flash 0.50
Position Consistency
6 models we track have a score here.
- 1 Kimi K2.5 0.63
- 2 GPT-5.2 ↑1 0.62
- 3
Qwen3.5-Plus ↓1 0.62
- 4 GLM-4.7 ↑1 0.61
- 5
DeepSeek-V4-Flash ↓1 0.60
- 6
Qwen3.5-Flash 0.53
Evidentiary Advocacy
6 models we track have a score here.
- 1 GPT-5.2 ↑2 0.61
- 2 Kimi K2.5 ↓1 0.55
- 3 GLM-4.7 ↑2 0.53
- 4
Qwen3.5-Plus ↓2 0.53
- 5
DeepSeek-V4-Flash ↓1 0.52
- 6
Qwen3.5-Flash 0.46
Legal Reasoning
6 models we track have a score here.
- 1 GPT-5.2 ↑2 0.57
- 2 Kimi K2.5 ↓1 0.56
- 3
Qwen3.5-Plus ↓1 0.53
- 4
DeepSeek-V4-Flash 0.52
- 5 GLM-4.7 0.48
- 6
Qwen3.5-Flash 0.45
Party Identification
6 models we track have a score here.
- 1 Kimi K2.5 0.85
- 2
Qwen3.5-Plus 0.85
- 3 GPT-5.2 0.83
- 4
DeepSeek-V4-Flash 0.83
- 5 GLM-4.7 0.82
- 6
Qwen3.5-Flash 0.82
Claim Construction
6 models we track have a score here.
- 1
Qwen3.5-Plus ↑1 0.72
- 2 Kimi K2.5 ↓1 0.71
- 3 GPT-5.2 0.70
- 4
Qwen3.5-Flash ↑2 0.69
- 5 GLM-4.7 0.66
- 6
DeepSeek-V4-Flash ↓2 0.65
Fact Marshalling
6 models we track have a score here.
- 1 Kimi K2.5 0.74
- 2
Qwen3.5-Flash ↑4 0.72
- 3 GLM-4.7 ↑2 0.70
- 4
Qwen3.5-Plus ↓2 0.70
- 5
DeepSeek-V4-Flash ↓1 0.69
- 6 GPT-5.2 ↓3 0.68
Evidence Marshalling
6 models we track have a score here.
- 1 GLM-4.7 ↑4 0.71
- 2
Qwen3.5-Plus 0.70
- 3 GPT-5.2 0.63
- 4
DeepSeek-V4-Flash 0.57
- 5
Qwen3.5-Flash ↑1 0.53
- 6 Kimi K2.5 ↓5 0.52
Position Consistency
6 models we track have a score here.
- 1 Kimi K2.5 0.65
- 2 GPT-5.2 ↑1 0.64
- 3
Qwen3.5-Plus ↓1 0.64
- 4
DeepSeek-V4-Flash 0.63
- 5 GLM-4.7 0.61
- 6
Qwen3.5-Flash 0.55
Evidentiary Advocacy
6 models we track have a score here.
- 1 Kimi K2.5 0.59
- 2 GPT-5.2 ↑1 0.55
- 3
Qwen3.5-Plus ↓1 0.54
- 4
DeepSeek-V4-Flash 0.54
- 5 GLM-4.7 0.52
- 6
Qwen3.5-Flash 0.41
Legal Reasoning
6 models we track have a score here.
- 1 Kimi K2.5 0.58
- 2
Qwen3.5-Plus 0.58
- 3 GPT-5.2 0.57
- 4
DeepSeek-V4-Flash 0.57
- 5 GLM-4.7 0.51
- 6
Qwen3.5-Flash 0.45
Position Consistency
6 models we track have a score here.
- 1 Kimi K2.5 0.64
- 2 GPT-5.2 ↑1 0.63
- 3
Qwen3.5-Plus ↓1 0.63
- 4
DeepSeek-V4-Flash 0.61
- 5 GLM-4.7 0.61
- 6
Qwen3.5-Flash 0.54
Evidentiary Advocacy
6 models we track have a score here.
- 1 GPT-5.2 ↑2 0.58
- 2 Kimi K2.5 ↓1 0.57
- 3
Qwen3.5-Plus ↓1 0.54
- 4
DeepSeek-V4-Flash 0.53
- 5 GLM-4.7 0.53
- 6
Qwen3.5-Flash 0.43
Legal Reasoning
6 models we track have a score here.
- 1 GPT-5.2 ↑2 0.57
- 2 Kimi K2.5 ↓1 0.57
- 3
Qwen3.5-Plus ↓1 0.56
- 4
DeepSeek-V4-Flash 0.55
- 5 GLM-4.7 0.49
- 6
Qwen3.5-Flash 0.45
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The LongJud-Bench paper
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official LongJud-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
- 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 7 months ago, in April 2026, for DeepSeek-V4-Flash. 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 (6 entries).
- Pass Models scored 6 models we track.
- Warning Newest score The newest came 7 months ago, in April 2026, for DeepSeek-V4-Flash.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.2 (#51).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 6 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.