Release Notes

BigLaw Bench: New Legal Task Evaluation Dataset

A new benchmark measures how well LLMs perform on real-world BigLaw tasks—Harvey leads in producing expert-level outputs.

Release Date

Sep 1, 2024

Categories

Model Updates

Harvey introduces the BigLaw Bench, a domain-specific dataset that simulates real-world legal tasks:

Based on billable legal work across transactional, litigation, and advisory contexts.
Harvey outperforms leading foundation models, producing 74% final-lawyer quality output—delivering nuanced, detailed work products that are close to ready for client delivery.