New Study: AI Legal Performance Is Converging on Routine Work, but Diverging Sharply on Complex Reasoning

General News

Summary

Percipient released a benchmark study showing that legal AI models are now close in performance on routine document work but still vary widely on complex legal reasoning tasks. The study evaluated frontier models from Anthropic, OpenAI, Google, xAI, Kimi, and DeepSeek across litigation, transactional, employment, and insurance workflows. It found that reasoning-focused models outperform standard models on more difficult analysis, especially where noise resistance and multi-step judgment matter. The report is aimed at helping law firms and legal departments evaluate AI tools more meaningfully for real legal work.

Classifications

industries
HealthTech
applications
Accounting and Taxes

AskAI Classifications

Labels
SaaS Consumer Software Enterprise Software

Linked Companies

Google LLC
$100M to $250M
xAI
$10M to $25M
OpenAI
$25M to $50M
Percipient
$5M to $10M
Anthropic
$10M to $25M