Harvey Integrates GPT-5.6 Sol — 92.7% on BigLaw Bench, Tokens Up 14x in Six Months

Harvey deployed GPT-5.6 Sol on July 9 for US customers (EU and AU availability to follow), achieving 92.7% on BigLaw Bench — a step up from GPT-5.5's 91.7% — with perfect scores in risk assessment, compliance, and deal management. The release accompanied co-founder Gabe Pereyra's disclosure that Harvey's token processing volume grew 14x in the past six months. A single contract review can now cost $20,000 in compute at the top-tier model level. Pereyra noted on X that passing token costs through to clients "might become more common — law firms already pass through costs for other legal tech, so not a completely crazy idea." For legal tech operators and law firm CFOs, the 14x usage growth combined with $20K-per-review compute costs frames the core commercial architecture problem of the next phase: Harvey is understood to be developing a new pricing model, and how it structures value vs. token pricing will set pricing norms across the sector. Harvey also launched 45+ new legal research sources, language localization for French, expanded vault sharing to iOS, and added reference files and Organization Context for in-house teams.