Legal AI's Token Price Problem Becomes a Structural Economics Issue
An Artificial Lawyer analysis published 3 June reports that rising frontier model pricing — driven by OpenAI and Anthropic moving away from subsidised per-seat models — is creating a spiralling cost problem for both legal AI vendors and law firms.
BY TRANSATLANTIC DESK · JUNE 3, 2026 · 1 MIN READ
An Artificial Lawyer analysis published 3 June reports that rising frontier model pricing — driven by OpenAI and Anthropic moving away from subsidised per-seat models — is creating a spiralling cost problem for both legal AI vendors and law firms. Agentic legal workflows are orders of magnitude more token-intensive than single-query interactions, and vendors absorbing the gap are under pressure to pass costs on, end flat-fee deals, or shift to consumption-based pricing. Harvey CEO Winston Weinberg confirmed the company is actively optimising token routing through Factory and LangChain to reduce agent evaluation costs. Thomson Reuters' investment in its own open-source model and Kirkland & Ellis's in-house LLM fine-tuning are both cited as signals that the economics of frontier-model dependency are unsustainable at scale.