Harvey AI
Review legal material and prepare drafts in the same workspace.
Harvey builds AI software for legal and professional services. Its platform supports legal workflows for firms and enterprise teams, combining application tools with domain-focused model development. The company was founded by lawyer Winston Weinberg and AI researcher Gabriel Pereyra.

Founded 2022 · San Francisco, CA, USA
Official WebsiteHarvey’s financing history follows its expansion from early legal AI tools to firm-wide workflows and agents that can handle several steps of legal work.
$550M · $15.5B valuation
The round followed Harvey’s release of its first post-trained open-weight model and the launch of its Legal Agent Benchmark.
Led by Diffusion, Lightspeed Venture Partners. Participation included Sequoia Capital, Kleiner Perkins, Andreessen Horowitz, Coatue, Conviction, Elad Gil, Evantic Capital, GIC, Goldman Sachs Alternatives, Verified Capital, WndrCo, Sapphire Ventures, Whale Rock Capital Management.
Source · harvey.ai ↗$200M · $11B valuation
Harvey said the new funding would expand the agents customers run on its platform and grow the legal engineering teams that help deploy them.
Led by GIC, Sequoia Capital. Participation included Andreessen Horowitz, Coatue, Conviction, Elad Gil, Evantic Capital, Kleiner Perkins.
Source · harvey.ai ↗$160M
Harvey announced the financing after a separate investment by EQT. It named WndrCo and accounts advised by T. Rowe Price as new backers.
Led by Andreessen Horowitz. Participation included WndrCo, T. Rowe Price, Sequoia Capital, Kleiner Perkins, Conviction, Elad Gil.
Source · harvey.ai ↗$300M · $5B valuation
Harvey announced a new financing, distinct from its Series D earlier that year.
Led by Kleiner Perkins, Coatue. Participation included Sequoia Capital, GV, DST Global, Conviction, Elad Gil, OpenAI Startup Fund, SV Angel.
Source · harvey.ai ↗US$300 million · $3B valuation
The announcement tied the financing to Harvey’s expansion of AI workflows for legal and professional-services teams.
Led by Sequoia Capital. Participation included Coatue, REV.
Source · harvey.ai ↗$100M · $1.5B valuation
Harvey said it would use the financing to improve its AI tools and expand them to more business functions and regions.
Led by GV. Participation included OpenAI, Kleiner Perkins, Sequoia Capital, Elad Gil, SV Angel.
Source · techcrunch.com ↗$80M · $715M valuation
Harvey planned to expand its work on custom models, add staff, and build out product features for law firms and other professional services teams.
Led by Elad Gil, Kleiner Perkins. Participation included OpenAI Startup Fund, Sequoia Capital.
Source · harvey.ai ↗$21M
The Series A backed Harvey’s initial focus on legal work as the starting point for applying generative AI across professional services.
Led by Sequoia Capital. Participation included OpenAI Startup Fund, Conviction, SV Angel, Elad Gil.
Source · harvey.ai ↗$5M
Harvey emerged from stealth with plans to build an AI assistant for lawyers.
Led by OpenAI Startup Fund. Participation included Jeff Dean, Elad Gil.
Source · techcrunch.com ↗Harvey’s founding team brought legal practice and AI research together to build tools for lawyers.
Review legal material and prepare drafts in the same workspace.
Harvey broadened its law-firm assistant into a platform for agents across law firms and in-house legal teams. Its 2026 announcement of the Guardrails AI acquisition and the separate Tenet research preview show work on two layers underneath those workflows: how agent behavior is controlled and how models handle longer legal tasks.
Harvey supports drafting and document analysis within law-firm workflows. CoCounsel combines legal assistance with Thomson Reuters’ research products, including Westlaw. For a legal team choosing between them, the relevant distinction is how the assistant fits the firm’s document work and research subscriptions.
Because the relationship predates the firm: Diffusion was co-founded by Kris Fredrickson, a longtime Harvey backer from his Coatue years — he was there for the June 2025 growth round Coatue co-led. The September 2026 syndicate is otherwise continuity — GIC, Sequoia, a16z, Kleiner Perkins, Conviction, Elad Gil all repeat — so the 'new' co-lead is a partner changing desks, not a new bet.
Because the moat moves down the stack: weeks before the September round, Harvey launched Tenet — an open-weight legal model post-trained on Moonshot's Kimi K3, a Chinese open base rather than a US one — with proceeds earmarked for model talent and compute. That ends dependence on OpenAI (its seed-era backer) and Anthropic for core capability, echoing Mistral's open-weight doctrine.
Because the acquisition it wanted collapsed: The Information reported Harvey pursuing ~US$600M at US$2B+ to buy legal-research service vLex; when that fell through, July 2024's Series C landed at US$100M and a US$1.5B valuation. The 2026 consolidation urge returned as tuck-ins — four acquisitions that year, capped by Guardrails AI at the US$15.6B round — teams bought from a far stronger base.
Mostly revenue doubling: US$11B (March 2026 Series G) to ~US$15.6B (September), with roughly 80% of top US law firms as customers. Lightspeed supplies the ceiling argument — legal as the second-largest addressable AI market after code — and code's number one shows the headroom: Cognition at US$48B on ~US$900M run-rate, about three times Harvey's mark at comparable growth.
Financially yes, strategically less each quarter: the OpenAI Startup Fund anchored the 2022 seed and followed through Series A, B and the 2024 Series C. But Tenet's Kimi K3 base and model-talent spending reposition Harvey from OpenAI-dependent to model-agnostic — a move Lovable paralleled with an in-house trained model, and Clay never faced, selling workflow to OpenAI as a customer.
A law firm may evaluate an assistant, purchase seats or embed it in document workflows; these are different levels of adoption. Harvey's financing history does not establish how each customer uses it. Separate named relationships from paid deployment, retention and verified legal-work outcomes.