Niko Grupen

Harvey Co-Founder Gabe Pereyra on the Token Pricing Reckoning Coming for AI

Gabe Pereyra, Niko Grupen, Molly O'Shea

Harvey Labs launched the open-source Legal Agent Benchmark (Lab) to evaluate AI performance on specialized legal tasks using synthetic data validated by human lawyers, replacing generic QA with rigorous, task-specific unit tests. Initial findings reveal a fragmented market where no single model dominates all domains, prompting a strategic shift toward diverse model routing and post-training open-weight solutions to navigate cost-quality tradeoffs. While usage has reached 13 trillion tokens, the industry now prioritizes organizational intelligence and billing transparency over raw capability, driven by the need to resolve conflict-of-interest risks and optimize complex agent workflows.