LexPilot
Models interview a user about a Chinese civil case and predict the ruling on each claim. Paired consultations reveal the effect of missing facts.
- Made by
- Harbin Institute of Technology and Tsinghua University
- Released
- October 2026
- Kind
- Law
- Level
- Expert
- Scored
- Graded by a model
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Subject
- Law
- Language
- Chinese
- Interaction
- Multi-turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Law and finance
- Strongest model tested
- Claude Opus 5 #11 on the capability index
- Last model added
- September 2026
- Scores refreshed
- November 20th, 2026
Claude Opus 5 leads with 72.1% among models released by September 2026 Claude Opus 5 leads with 72.1% among models released by September 2026
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 reports separate consultation tasks. The headline is macro F1 for Legal Completeness Assessment, expressed as a percentage. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
The newest model with a score here came out in September 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 14 more
- 1 Claude Opus 5 72.1%
- 2 GPT-5.6 Sol 70.7%
- 3
Gemini 3.8 Flash 70.4%
- 4 LegalOne-8B 67.7%
- 5 GLM-5.3 65%
- 6 Kimi K3 64%
- 7
MiniMax-M3 63.3%
- 8
DeepSeek-V4-Pro 61.5%
- 9
Qwen 3.8 Max 59.6%
- 10 LawLLM-7B 41.9%
Source: The LexPilot paper
- 1 Claude Opus 5 72.1%
- 2 GPT-5.6 Sol 70.7%
- 3
Gemini 3.8 Flash 70.4%
- 4 GLM-5.3 65%
- 5 Kimi K3 64%
- 6
MiniMax-M3 63.3%
- 7
DeepSeek-V4-Pro 61.5%
- 8
Qwen 3.8 Max 59.6%
Source: The LexPilot paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 5 | 72.1% | Jul 2026 | Official leaderboard | |
| 2 | GPT-5.6 Sol | 70.7% | Jul 2026 | Official leaderboard | |
| 3 | | 70.4% | Sep 2026 | Official leaderboard | |
| 4 | LegalOne-8B | 67.7% | Official leaderboard | ||
| 5 | GLM-5.3 | 65% | Aug 2026 | Official leaderboard | |
| 6 | Kimi K3 | 64% | Jul 2026 | Official leaderboard | |
| 7 | | 63.3% | Jun 2026 | Official leaderboard | |
| 8 | | 61.5% | Apr 2026 | Official leaderboard | |
| 9 | | 59.6% | Aug 2026 | Official leaderboard | |
| 10 | LawLLM-7B | 41.9% | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 5 | 72.1% | Jul 2026 | Official leaderboard | |
| 2 | GPT-5.6 Sol | 70.7% | Jul 2026 | Official leaderboard | |
| 3 | | 70.4% | Sep 2026 | Official leaderboard | |
| 4 | GLM-5.3 | 65% | Aug 2026 | Official leaderboard | |
| 5 | Kimi K3 | 64% | Jul 2026 | Official leaderboard | |
| 6 | | 63.3% | Jun 2026 | Official leaderboard | |
| 7 | | 61.5% | Apr 2026 | Official leaderboard | |
| 8 | | 59.6% | Aug 2026 | Official leaderboard |
Rankings by part
We read 29 parts from the benchmark’s paper in 4 groups: legal completeness assessment, detail completion guidance, interaction-based ljp and gold-standard ljp. Each ranks models on one part of the test. Claude Opus 5 leads 16 of them, and 4 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Direct LJP
Quadratic weighted kappa. 8 models we track have a score here.
- 1 Claude Opus 5 0.43
- 2 GPT-5.6 Sol 0.43
- 3 GLM-5.3 ↑1 0.38
- 4
MiniMax-M3 ↑2 0.38
- 5
Gemini 3.8 Flash ↓2 0.36
- 6
DeepSeek-V4-Pro ↑1 0.35
- 7
Qwen 3.8 Max ↑1 0.35
- 8 Kimi K3 ↓3 0.32
Legal Completeness Assessment
8 models we track have a score here.
- 1 Claude Opus 5 72.1%
- 2 GPT-5.6 Sol 70.7%
- 3
Gemini 3.8 Flash 70.4%
- 4 GLM-5.3 65%
- 5 Kimi K3 64%
- 6
MiniMax-M3 63.3%
- 7
DeepSeek-V4-Pro 61.5%
- 8
Qwen 3.8 Max 59.6%
Detail Completion Guidance
8 models we track have a score here.
- 1 Claude Opus 5 74.2%
- 2 GPT-5.6 Sol 73.4%
- 3 GLM-5.3 ↑1 69.2%
- 4
Gemini 3.8 Flash ↓1 68.9%
- 5
Qwen 3.8 Max ↑3 67.9%
- 6
DeepSeek-V4-Pro ↑1 66.4%
- 7 Kimi K3 ↓2 66.2%
- 8
MiniMax-M3 ↓2 61.1%
Interaction-Based LJP
8 models we track have a score here.
- 1 Claude Opus 5 0.52
- 2 GPT-5.6 Sol 0.48
- 3 GLM-5.3 ↑1 0.44
- 4
Gemini 3.8 Flash ↓1 0.43
- 5
DeepSeek-V4-Pro ↑2 0.40
- 6
Qwen 3.8 Max ↑2 0.39
- 7
MiniMax-M3 ↓1 0.37
- 8 Kimi K3 ↓3 0.37
Gold-Standard LJP
8 models we track have a score here.
- 1 Claude Opus 5 0.65
- 2 GPT-5.6 Sol 0.62
- 3
Gemini 3.8 Flash 0.59
- 4 GLM-5.3 0.55
- 5
MiniMax-M3 ↑1 0.52
- 6
DeepSeek-V4-Pro ↑1 0.50
- 7 Kimi K3 ↓2 0.48
- 8
Qwen 3.8 Max 0.47
Family
8 models we track have a score here.
- 1 Claude Opus 5 75%
- 2 GPT-5.6 Sol 73.5%
- 3
Gemini 3.8 Flash 72.2%
- 4
DeepSeek-V4-Pro ↑3 66.7%
- 5 GLM-5.3 ↓1 66.2%
- 6
MiniMax-M3 65.7%
- 7 Kimi K3 ↓2 65.6%
- 8
Qwen 3.8 Max 55%
Real Rights
8 models we track have a score here.
- 1 Claude Opus 5 76.3%
- 2 GPT-5.6 Sol 73.8%
- 3
Gemini 3.8 Flash 73.8%
- 4
MiniMax-M3 ↑2 67%
- 5 Kimi K3 66.4%
- 6
DeepSeek-V4-Pro ↑1 64.7%
- 7
Qwen 3.8 Max ↑1 61.5%
- 8 GLM-5.3 ↓4 61.1%
Torts
8 models we track have a score here.
- 1 Claude Opus 5 71%
- 2 Kimi K3 ↑3 70.9%
- 3 GPT-5.6 Sol ↓1 70.3%
- 4
Gemini 3.8 Flash ↓1 66.7%
- 5 GLM-5.3 ↓1 66.1%
- 6
DeepSeek-V4-Pro ↑1 65.5%
- 7
MiniMax-M3 ↓1 58.8%
- 8
Qwen 3.8 Max 58.3%
Service Contract
8 models we track have a score here.
- 1 GPT-5.6 Sol ↑1 70.7%
- 2
Gemini 3.8 Flash ↑1 70.5%
- 3 Claude Opus 5 ↓2 68%
- 4 GLM-5.3 65.2%
- 5
MiniMax-M3 ↑1 57.9%
- 6
DeepSeek-V4-Pro ↑1 57%
- 7
Qwen 3.8 Max ↑1 52.7%
- 8 Kimi K3 ↓3 51.9%
Finance Contract
8 models we track have a score here.
- 1 Claude Opus 5 72.3%
- 2
Gemini 3.8 Flash ↑1 69%
- 3 GPT-5.6 Sol ↓1 68%
- 4
MiniMax-M3 ↑2 64.3%
- 5 GLM-5.3 ↓1 64.2%
- 6
Qwen 3.8 Max ↑2 64.2%
- 7 Kimi K3 ↓2 62.6%
- 8
DeepSeek-V4-Pro ↓1 59.5%
Misc.
8 models we track have a score here.
- 1 Kimi K3 ↑4 69.3%
- 2 GLM-5.3 ↑2 69.3%
- 3 Claude Opus 5 ↓2 68.3%
- 4 GPT-5.6 Sol ↓2 67.2%
- 5
Gemini 3.8 Flash ↓2 67.2%
- 6
MiniMax-M3 64.4%
- 7
Qwen 3.8 Max ↑1 64%
- 8
DeepSeek-V4-Pro ↓1 56.2%
Family
8 models we track have a score here.
- 1 GPT-5.6 Sol ↑1 83.7%
- 2
Gemini 3.8 Flash ↑1 78%
- 3 Claude Opus 5 ↓2 75.3%
- 4 GLM-5.3 74.7%
- 5 Kimi K3 70.7%
- 6
DeepSeek-V4-Pro ↑1 67.3%
- 7
MiniMax-M3 ↓1 67.3%
- 8
Qwen 3.8 Max 65.3%
Real Rights
8 models we track have a score here.
- 1 Claude Opus 5 80.8%
- 2 GPT-5.6 Sol 78.8%
- 3 Kimi K3 ↑2 75.4%
- 4 GLM-5.3 75.4%
- 5
DeepSeek-V4-Pro ↑2 75.2%
- 6
Qwen 3.8 Max ↑2 74.4%
- 7
Gemini 3.8 Flash ↓4 72.3%
- 8
MiniMax-M3 ↓2 66%
Torts
8 models we track have a score here.
- 1 Kimi K3 ↑4 77.5%
- 2 Claude Opus 5 ↓1 77.5%
- 3
Qwen 3.8 Max ↑5 77.1%
- 4 GPT-5.6 Sol ↓2 76.3%
- 5
DeepSeek-V4-Pro ↑2 75%
- 6
MiniMax-M3 75%
- 7 GLM-5.3 ↓3 71.7%
- 8
Gemini 3.8 Flash ↓5 69.6%
Service Contract
8 models we track have a score here.
- 1 GPT-5.6 Sol ↑1 74%
- 2
Qwen 3.8 Max ↑6 66.7%
- 3 Claude Opus 5 ↓2 66%
- 4
Gemini 3.8 Flash ↓1 64%
- 5
DeepSeek-V4-Pro ↑2 62.1%
- 6 GLM-5.3 ↓2 62.1%
- 7 Kimi K3 ↓2 52.4%
- 8
MiniMax-M3 ↓2 51.2%
Finance Contract
8 models we track have a score here.
- 1 Claude Opus 5 69%
- 2 GLM-5.3 ↑2 68.8%
- 3
Gemini 3.8 Flash 68.3%
- 4 Kimi K3 ↑1 61.4%
- 5 GPT-5.6 Sol ↓3 60.7%
- 6
Qwen 3.8 Max ↑2 59.8%
- 7
MiniMax-M3 ↓1 53.8%
- 8
DeepSeek-V4-Pro ↓1 53.6%
Misc.
8 models we track have a score here.
- 1 Claude Opus 5 78.3%
- 2 GPT-5.6 Sol 69.3%
- 3
DeepSeek-V4-Pro ↑4 68.7%
- 4
Qwen 3.8 Max ↑4 65.7%
- 5 Kimi K3 64%
- 6 GLM-5.3 ↓2 62%
- 7
Gemini 3.8 Flash ↓4 61.7%
- 8
MiniMax-M3 ↓2 59.7%
Family
8 models we track have a score here.
- 1 GPT-5.6 Sol ↑1 0.60
- 2 Kimi K3 ↑3 0.55
- 3
Gemini 3.8 Flash 0.54
- 4 GLM-5.3 0.54
- 5 Claude Opus 5 ↓4 0.50
- 6
Qwen 3.8 Max ↑2 0.49
- 7
DeepSeek-V4-Pro 0.48
- 8
MiniMax-M3 ↓2 0.44
Real Rights
8 models we track have a score here.
- 1 GLM-5.3 ↑3 0.52
- 2 Claude Opus 5 ↓1 0.49
- 3
Qwen 3.8 Max ↑5 0.46
- 4
MiniMax-M3 ↑2 0.42
- 5 GPT-5.6 Sol ↓3 0.40
- 6
Gemini 3.8 Flash ↓3 0.36
- 7
DeepSeek-V4-Pro 0.35
- 8 Kimi K3 ↓3 0.25
Torts
8 models we track have a score here.
- 1 GLM-5.3 ↑3 0.53
- 2 GPT-5.6 Sol 0.50
- 3 Claude Opus 5 ↓2 0.48
- 4
Qwen 3.8 Max ↑4 0.46
- 5
MiniMax-M3 ↑1 0.44
- 6
DeepSeek-V4-Pro ↑1 0.44
- 7 Kimi K3 ↓2 0.39
- 8
Gemini 3.8 Flash ↓5 0.30
Service Contract
8 models we track have a score here.
- 1 Claude Opus 5 0.56
- 2
Gemini 3.8 Flash ↑1 0.50
- 3 Kimi K3 ↑2 0.42
- 4 GPT-5.6 Sol ↓2 0.41
- 5
DeepSeek-V4-Pro ↑2 0.40
- 6 GLM-5.3 ↓2 0.35
- 7
Qwen 3.8 Max ↑1 0.33
- 8
MiniMax-M3 ↓2 0.33
Finance Contract
8 models we track have a score here.
- 1 Claude Opus 5 0.42
- 2 GPT-5.6 Sol 0.24
- 3
Qwen 3.8 Max ↑5 0.19
- 4
Gemini 3.8 Flash ↓1 0.16
- 5 Kimi K3 0.14
- 6 GLM-5.3 ↓2 0.12
- 7
DeepSeek-V4-Pro 0.11
- 8
MiniMax-M3 ↓2 0.07
Misc.
8 models we track have a score here.
- 1 GPT-5.6 Sol ↑1 0.67
- 2 GLM-5.3 ↑2 0.63
- 3 Claude Opus 5 ↓2 0.62
- 4
Gemini 3.8 Flash ↓1 0.61
- 5
DeepSeek-V4-Pro ↑2 0.60
- 6 Kimi K3 ↓1 0.50
- 7
MiniMax-M3 ↓1 0.48
- 8
Qwen 3.8 Max 0.41
Family
8 models we track have a score here.
- 1 GPT-5.6 Sol ↑1 0.74
- 2 Claude Opus 5 ↓1 0.73
- 3
MiniMax-M3 ↑3 0.65
- 4
Qwen 3.8 Max ↑4 0.64
- 5 GLM-5.3 ↓1 0.62
- 6
DeepSeek-V4-Pro ↑1 0.61
- 7
Gemini 3.8 Flash ↓4 0.60
- 8 Kimi K3 ↓3 0.59
Real Rights
8 models we track have a score here.
- 1 Claude Opus 5 0.68
- 2 GPT-5.6 Sol 0.58
- 3
Gemini 3.8 Flash 0.57
- 4 Kimi K3 ↑1 0.57
- 5
DeepSeek-V4-Pro ↑2 0.56
- 6 GLM-5.3 ↓2 0.52
- 7
MiniMax-M3 ↓1 0.48
- 8
Qwen 3.8 Max 0.45
Torts
8 models we track have a score here.
- 1 GLM-5.3 ↑3 0.60
- 2 GPT-5.6 Sol 0.57
- 3 Claude Opus 5 ↓2 0.53
- 4
Gemini 3.8 Flash ↓1 0.51
- 5
Qwen 3.8 Max ↑3 0.46
- 6
MiniMax-M3 0.43
- 7 Kimi K3 ↓2 0.38
- 8
DeepSeek-V4-Pro ↓1 0.29
Service Contract
8 models we track have a score here.
- 1 GPT-5.6 Sol ↑1 0.72
- 2 Claude Opus 5 ↓1 0.66
- 3
MiniMax-M3 ↑3 0.65
- 4
Gemini 3.8 Flash ↓1 0.62
- 5 Kimi K3 0.50
- 6
DeepSeek-V4-Pro ↑1 0.47
- 7 GLM-5.3 ↓3 0.46
- 8
Qwen 3.8 Max 0.40
Finance Contract
8 models we track have a score here.
- 1 Claude Opus 5 0.50
- 2
Gemini 3.8 Flash ↑1 0.39
- 3
DeepSeek-V4-Pro ↑4 0.38
- 4 GLM-5.3 0.37
- 5 GPT-5.6 Sol ↓3 0.33
- 6
Qwen 3.8 Max ↑2 0.29
- 7 Kimi K3 ↓2 0.21
- 8
MiniMax-M3 ↓2 0.17
Misc.
8 models we track have a score here.
- 1
Gemini 3.8 Flash ↑2 0.77
- 2 GLM-5.3 ↑2 0.73
- 3 Claude Opus 5 ↓2 0.69
- 4 GPT-5.6 Sol ↓2 0.68
- 5
MiniMax-M3 ↑1 0.60
- 6
DeepSeek-V4-Pro ↑1 0.60
- 7 Kimi K3 ↓2 0.55
- 8
Qwen 3.8 Max 0.51
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The LexPilot 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 LexPilot 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
All 10 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the benchmark’s paper (10 entries).
- Pass Models scored 8 models we track, plus 2 other entries on its leaderboard.
- Pass Newest score The newest is from September 2026, for Gemini 3.8 Flash.
- Note Frontier models None of the top 10 has a score. The strongest tested is Claude Opus 5 (#11).
- Pass Headroom The best is 72.1%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 8 rows for models we track and 2 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.