Harvey raises $550 million at $15.5 billion for legal AI

Harvey's raise comes just six months after its previous nine-figure funding round, a quick cadence that suggests investors are betting on legal-specific AI rather than general-purpose models. The company says its tools are already in use at 80% of the 100 highest-ranked U.S. law firms and half the Fortune 10, a claim that positions it as an incumbent among enterprise legal teams rather than a challenger.

The new capital follows the launch of Tenet, Harvey's first custom large language model. Tenet is a fine-tuned version of Kimi K3, an open-source model with 2.8 trillion parameters, and Harvey says it performs some contract processing tasks 20% better than its base. Harvey also debuted LAB, a benchmark for legal work that currently includes about 1,200 tasks. The company's stated plan is to deploy more computing infrastructure and prioritize 'new generalist models.'

Developing proprietary models is expensive upfront but can lower long-term infrastructure costs by reducing reliance on external models, since inference typically accounts for a larger share of AI workload cost than training. Whether Harvey's margins improve will hinge on how quickly it can shift workloads onto Tenet and future models, and on whether its LAB benchmark gains adoption beyond its current scope.