reasoning benchmark
Legal Agent Benchmark
The Legal Agent Benchmark (LAB) is Harvey's open-source benchmark for evaluating AI agents on complex, long-horizon legal work. Tasks are scored under an all-pass standard against expert-curated rubrics, where a task passes only if every required rubric criterion (facts, conclusions, citations, structure, and analytical moves) passes.
Updated Aug 7, 2026
Published models13
Registry coverage13
MetricScore
EvidenceB
Legal Agent Benchmark leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | AN | 13.3% | 100.0% | 13 | C | |
| 2 | AN | 11.7% | 91.7% | 13 | C | |
| 3 | AN | 7.1% | 83.3% | 13 | B | |
| 4 | AN | 5.8% | 75.0% | 13 | C | |
| 5 | AN | 5.4% | 66.7% | 13 | B | |
| 6 | AN | 4.2% | 58.3% | 13 | B | |
| 7 | OP | 2.1% | 50.0% | 13 | B | |
| 8 | GO | 0.8% | 41.7% | 13 | B | |
| 9 | OP | 0.4% | 33.3% | 13 | B | |
| 10 | GO | 0.0% | 25.0% | 13 | B | |
| 11 | GO | 0.0% | 16.7% | 13 | B | |
| 12 | GO | 0.0% | 8.3% | 13 | B | |
| 13 | OP | 0.0% | 0.0% | 13 | B |
Score distribution
Top published rows on the benchmark's original scale.
Legal Agent Benchmark highlights
The top published results on this benchmark's own scale.
What is Legal Agent Benchmark?
Definition and scoring fields from the benchmark registry.
The Legal Agent Benchmark (LAB) is Harvey's open-source benchmark for evaluating AI agents on complex, long-horizon legal work. Tasks are scored under an all-pass standard against expert-curated rubrics, where a task passes only if every required rubric criterion (facts, conclusions, citations, structure, and analytical moves) passes.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- Legal Agent Benchmark
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- legal-agent-benchmark|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about Legal Agent Benchmark.
Which model scores highest on Legal Agent Benchmark?
Claude Fable 5 is currently ranked first with 13.3%.
What does Legal Agent Benchmark measure?
The Legal Agent Benchmark (LAB) is Harvey's open-source benchmark for evaluating AI agents on complex, long-horizon legal work. Tasks are scored under an all-pass standard against expert-curated rubrics, where a task passes only if every required rubric criterion (facts, conclusions, citations, structure, and analytical moves) passes.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
13 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.