Harvey AI raised $200 million last week at an $11 billion valuation. What it didn't announce: it abandoned the custom legal model it was built on.
The pivot is the real story buried under the funding headline. Harvey was constructed around a proprietary vertical model — a custom-trained system designed to outperform general-purpose language models on legal reasoning tasks. Then frontier models got better. Specifically, according to research from analytics firm Sacra, reasoning models from Google, xAI, OpenAI, and Anthropic began outperforming Harvey's own custom legal model on Harvey's own benchmark, called BigLaw Bench. Harvey scrapped the proprietary approach. It now runs multi-LLM orchestration — chaining different models and tools together depending on the task. "AI isn't just assisting lawyers. It's becoming the system through which legal work gets done," Weinberg told GIC's press release. That framing is technically accurate. It is also, in part, a retreat narrative dressed as a platform story.
The funding context makes the pivot worth flagging. Singapore's sovereign wealth fund GIC has made AI a declared strategic investment focus — it co-led the $200 million round alongside Sequoia Capital, which has now led three of Harvey's funding rounds. That positioning matters: GIC previously co-led Anthropic's $30 billion Series G and has explicitly tied AI to its long-term portfolio transformation goals. The Harvey bet fits that thesis, not a departure from it. At $190 million in annual recurring revenue as of January 2026, per CNBC, the company carries an approximately 58x revenue multiple — a price that assumes the revenue is durable, the market share is defensible, and the platform story holds. Whether any of those assumptions survive contact with reality is the open question.
The numbers that are publicly available are substantial. Harvey says it has more than 25,000 custom agents operating on its platform, serving more than half of the AmLaw 100, over 500 in-house legal teams, and 100,000 lawyers across 1,300 organizations in 60 countries. Revenue grew from $50 million ARR at the end of 2024 to $100 million by August 2025 and $190 million by January 2026 — nearly quadrupling in twelve months, according to Sacra, whose independent estimate of $195 million end-2025 ARR tracks closely with Harvey's self-reported figure. The growth rate is real. The self-reported nature of all of it should temper confidence in the precision.
One tension worth naming: Paul Weiss, a major law firm, told Business Insider it is not measuring whether Harvey saves time. The structural reason for that is its own story. Big Law runs on billable hours. There is a fine line between AI streamlining how lawyers work and cutting into how much they can charge clients. If Harvey genuinely reduces legal work volume, it threatens the revenue model of the firms it sells to. Nobody in the funding coverage is asking this question directly.
Harvey's pricing — $1,200 per lawyer per month with 12-month commitments and roughly 20-seat minimums — means a single firm commitment runs $288,000 annually at base. That is contracted revenue, not a pilot. But it also means Harvey's customer base is the upper end of the legal market. Whether the platform extends down to mid-market and solo practitioners, or whether it consolidates around a small number of large firms running deep agent deployments, is not yet answered.
The GIC angle deserves its own consideration. GIC CEO Lim Chow Kiat explicitly tied AI to long-term transformation goals at a July 2025 press conference, per Undercode News. The fund has flagged AI as a strategic investment focus and increased its U.S. portfolio exposure to 49 percent. GIC also co-led Anthropic's $30 billion Series G at a $380 billion post-money valuation. Buying into Harvey at $11 billion is a different kind of bet — not frontier model infrastructure, but vertical deployment and distribution. The pattern suggests GIC is building a layer of exposure to how AI actually gets used inside enterprises, not just who builds the base models.
Harvey has made concrete moves to support the platform framing. It acquired Hexus in January 2026 — a company building AI tools for creating product demos and guides, positioned to extend Harvey's reach into in-house legal teams. It entered a strategic alliance with LexisNexis in June 2025, integrating Protégé AI, primary law content, and Shepard's Citations — foundational legal content that is genuinely hard for a startup to build or license from scratch. Pat Grady of Sequoia, speaking in the GIC press release, put the strongest version of the platform claim: "Harvey has become the platform on which legal work runs."
The competitive picture complicates that claim. Thomson Reuters acquired Casetext for $650 million in 2023 and has been building its own AI legal layer. A Thomson Reuters VP of AI described the market as bifurcating into operational AI — where Harvey sits — and authoritative AI, where the incumbents live. Anthropic's Cowork plugin specifically attacks operational legal work with enterprise connectors and custom plugins, positioning itself as a potential operating system for legal operations. Harvey sits between incumbents with authoritative infrastructure and foundation model companies with enormous scale advantages. That is an increasingly difficult strategic position, as Fortune noted.
What to watch: whether Harvey uses this capital to move beyond English-language common-law markets, where it has deepest traction, into civil law jurisdictions and international firms. The 60 countries in the coverage is broad but thin. The real measure of platform durability is whether those 100,000 lawyers are deepening their use of Harvey's agents or treating it as one vendor among several. If GIC's bet is right, that deepening is already happening. If it is wrong, the next round gets harder to justify.
Harvey has more than $1 billion in total funding, according to its own announcement. It has the capital to find out.
