LB

Harvey AI Legal Agent Benchmark

Browse 1,182 real legal tasks across 24 practice areas.

Every task in the open harvey-labs benchmark, drillable to the prompt, full rubric, and reference documents the model is expected to consult.

Tasks
1,182
1,251 scenarios
Practice areas
24
legal specialties
Rubric criteria
74,990
avg 60 per task
Reference docs
9,527
.docx, .xlsx, .pdf, .eml

Work type breakdown

Across all 1,251 scenarios

analyze 490 draft 444 review 293 research 24

Largest practice areas

Top 9 of 24, ranked by number of tasks.

View all 24 →