Harvey vs Legora: The 35-Hour Onboarding vs the 6-Month Build

Harvey and Legora each closed a new funding round this autumn, six weeks apart. Harvey's, announced on 9 September 2026, valued it at $15.5 billion. Legora's Series E, announced on 21 October, valued it at $8.8 billion post-money.

So the valuation gap is real but narrowing: on paper Harvey is now worth about 1.8 times Legora, down from almost three times in September. Neither publishes a price list. The number that should decide a firm's 2027 legal AI budget is not that gap, though. It is how long each platform takes to work inside a law firm, and what a seat costs once the minimums kick in.

Hokai comparison of how Harvey and Legora get into a firm and what the minimums cost, from the funding posts, case studies and reports the guide cites.

Legora's own announcement of its March 2026 Series D, the round before the October Series E, captured 30 Oct 2026.

What Changed Since 16 August 2026

This comparison was first published on 16 August 2026 and re-checked against both vendors' own pages on 12 October and again on 6 November 2026. The lineup holds: these are still the two AI-native platforms competing for the same large-firm budgets, and the verdict below is unchanged. Four facts moved.

• Harvey's round closed. The talks reported in August became a signed $550 million round on 9 September.
• Legora's round closed below the reported target. A report in August put its target at $10 billion or more. On 21 October it announced a $550 million Series E at $8.8 billion post-money, led by Atreides Management.
• Harvey started training its own model. In August it released a research preview of a model post-trained from an open-weight base.
• Legora rebuilt legal research. On 14 September it described rebuilding its research product around a structured map of the law and a citator designed for AI, built partly with technology from two acquisitions.

Challengers considered and left out: Thomson Reuters' CoCounsel, which is covered below as the incumbent fallback; RealPact, which automates contracts for real estate brokerages, a different job; and general assistants, which get their own section.

Valuation and Revenue, Side by Side

Harvey's latest valuation is $15.5 billion, set on 9 September 2026, and Legora's is $8.8 billion, set on 21 October 2026. Both figures come from rounds the companies announced themselves.

• Harvey$15.5B Leader
• Legora$8.8B

Read from each company's own funding announcement.

The figures for Harvey come from its own funding posts of 25 March and 9 September, plus a PYMNTS report of 7 August that put its annualized revenue above $350 million, up from $190 million in January. It now says 80% of Am Law 100 firms use it, alongside five of the Fortune 10.

The Legora figures come from its own Series E release of 21 October, which also says it has more than 1,000 employees across 18 offices and that in-house legal teams made up more than 40% of its new customers in the third quarter. The April mark comes from TechCrunch's report on the Series D extension.

The two are separate, competing companies with no shared ownership: Harvey is American, while Legora is Swedish, was founded in 2023 and now names the United States as its largest market by revenue.

On every one of those lines Harvey is still the bigger company, though the gap closed quickly between September and October. The question for a buyer is whether bigger means better for their firm, and that is where the answer splits.

Pick Legora for Speed, Harvey to Build

For a firm under roughly 800 lawyers with a deadline to show partners that AI spending is working, Legora is the safer buy. Its German client GÖRG, a firm of more than 380 lawyers, ran 35 hours of face-to-face onboarding and reached 96.81% weekly active usage, according to a case study the vendor published. One attorney there logged 459 prompts in under three weeks.

That is a fast, measurable result a legal tech manager can report to a partner committee before the fiscal year closes. It is also a vendor case study, so treat it as the best case rather than the median.

Harvey's platform is the better buy for a firm that already has top-tier budget and an internal team willing to co-build. It embeds a forward deployed engineer, or FDE, full-time inside a client for six to nine months to turn the firm's own workflows into custom applications, according to a deployment breakdown from Perspective AI. That is a real commitment of time and headcount before the tool produces firm-specific value.

Neither answer is "it depends." Pick based on how much runway the firm can spend before it needs a result. If your shortlist is wider than these two, describe your firm in a few answers and see which legal tools fit before booking either demo.

What Each One Actually Costs

Neither vendor publishes a price, so every figure here is a third-party estimate and should be treated as an opening position until a real quote arrives. The direct answer: Legora costs far less at the contract floor, roughly $30,000 a year against roughly $360,000 for Harvey.

The per-seat range and seat minimum for Harvey come from a pricing breakdown by Vaquill, which cites analysts tracking leaked quotes. Costbench, reading four buyer reports and last verified on 1 July 2026, puts the median contract at $175,000 a year. The two do not contradict each other: the median buyer negotiated below the floor that the leaked quotes imply.

The list price, seat minimum and enterprise range for Legora come from a pricing analysis published on 11 May 2026 by Law XY AI. Vaquill reports a higher band for it, $300 to $800 per seat per month for unlimited use, and notes that it moved its agent tier to consumption-based pricing in June 2026.

• Harvey also sells a credit-based metered plan, per Vaquill, for teams that want to pay per use.
• Legora's metered agent tier suits firms with uneven agent workloads.
• Both add implementation, training and renewal uplift on top of the seat, which Vaquill puts at 30 to 50 percent above the headline license.

That gap, roughly 12 times at the contract floor, is the real starting point for a budget conversation. It matters more than either company's valuation. The plan detail for each vendor lives on the Harvey listing and the Legora listing.

How Each One Gets Into a Firm

The pricing gap tracks a real difference in how each platform gets built into a firm. The Perspective AI analysis describes the embedded program in three overlapping phases:

1. Firm discovery, weeks one to six, with 40 to 80 hours of interviews.
2. Workflow mapping, weeks four to fourteen.
3. Knowledge-base construction, weeks eight to twenty-eight.

The same analysis estimates there were 40 to 80 such engineers as of early 2026. The company did not publish that headcount; it is an outside estimate.

The output is custom agents. The March funding post said more than 25,000 custom agents run on the platform, across mergers, due diligence, contract drafting and document review. For a background read on the product itself, see what Harvey is for.

The Swedish company skips the embed. The legal tech manager who ran the GÖRG rollout said onboarding "should go beyond standard online sessions" and happen face to face. Nine months of bespoke engineering against 35 hours of training is the actual product difference here.

Who Else Is Betting on Each Platform

The investor lists tell a story neither pricing page does. OpenAI took part in an early $80 million round, when the company was building legal assistants on GPT-4 series models, according to coverage of that round by AIbase. The September round added four new investors to a cap table that already held Sequoia and Kleiner Perkins.

The March Series D for Legora was led by Accel. In April, the venture arm of Nvidia joined a $50 million extension, which TechCrunch reported as the chipmaker's first legal AI investment.

For a buyer, the practical read is about model strategy. The August research preview, called Tenet, is post-trained from a Kimi K3 base built by Moonshot AI, and the same research tested a GLM-5.2 orchestrator from Z.ai. You can compare those open-weight models on price and benchmark scores on the model leaderboard. The bigger vendor is moving toward owning its legal model rather than renting one lab's. The smaller one is putting its effort into legal data and a citator instead.

Where Harvey Wins

It wins for the largest firms running high-volume due diligence and litigation drafting that need deep document-management integration and in-region hosting. The March post put its reach at more than 100,000 lawyers across 1,300 organizations, and the September post says 80% of the Am Law 100 use it.

It also wins for firms that want to build bespoke agents instead of using templates. More than 25,000 custom agents is a scale of self-built automation no rival has disclosed. And it wins for firms that want their own model: the security page says a firm can ask for a bespoke model trained on its own data and used by no other customer.

Where Legora Wins

It wins for firms running cross-border or multilingual matters where collaboration between outside counsel and in-house teams matters more than deep automation. Its October release counts more than 2,400 law firms and in-house legal teams across over 80 markets, up from 1,000 in 50 markets at the April round, and several large companies, BMW Group among them, have announced rollouts on its newsroom.

It also wins for a mid-size firm that needs to prove return on investment fast, as the GÖRG numbers show. Its growth backs up the speed argument. The company grew from 40 to 400 people in the year to March 2026, according to its Series D release, and passed 1,000 employees across 18 offices by its Series E release in October.

On 28 September it opened a Munich office, saying its number of users in Germany had quadrupled. That matters to a firm in that market comparing local support.

Data Handling and Security

Both vendors make the commitment a law firm needs first: client data is not used to train models. The detail behind that promise differs, and so do the controls.

Every row comes from the two vendors' own security pages, read on 12 October 2026. The larger one runs its cloud on Microsoft Azure and offers in-region processing that, per its page, also binds its subprocessors. The smaller one's stand-out control is bring-your-own-key encryption, which lets a firm hold the keys to its own data.

Neither page names every model provider behind the product. A firm with a strict outside-counsel guideline should ask both vendors for the subprocessor list before signing, and get the answer in the contract rather than in a sales email.

Where General Assistants Like Claude Fit

A frequent follow-up is whether a firm needs either startup at all, or whether a general assistant will do. When Anthropic launched a legal plug-in for its assistant, several listed legal software companies saw their shares drop, TechCrunch noted in April.

A team seat on Claude or ChatGPT costs a small fraction of either legal platform. What it does not include is the legal layer: in-region hosting and custom agents on one side, a citator and tabular review across thousands of documents on the other. For a firm still deciding what a general assistant can carry, our guide to choosing an AI assistant and what Claude is good for are the place to start. The full field of legal tools sits in the AI legal assistant directory.

How to Run a Fair Trial

A demo shows what a vendor wants you to see. A trial shows what your lawyers will actually do with the tool, and the two rarely match. If the budget allows a paid pilot of both, run them the same way so the result means something.

1. Pick three real matters that closed in the last year, with the work product kept, so the output can be checked against what your team actually filed.
2. Give both tools the same documents, the same instructions and the same deadline.
3. Have one senior lawyer grade every answer blind, for accuracy and citations, without knowing which tool produced it.
4. Track weekly active users, not logins, for at least four weeks.
5. Ask each vendor for the price of the seat count you would actually buy, in writing.

The usage number is the one to watch. A tool that wins on accuracy but sits unused after the first week will not survive a renewal, and a tool lawyers open every day will find its way into more work over time. The Legora profile and the matching listing for its rival both record how each sells, so you can ask the right pilot questions up front.

Keep the pilot short and the scope narrow. Two practice groups are enough to see the pattern, and a pilot that tries to cover the whole firm usually ends without a clear answer.

Write the success test down before the pilot starts. If the team agrees in advance what usage rate and what accuracy grade would justify the spend, the decision at the end takes one meeting instead of a quarter of debate.

A Hot Valuation Is Not a Feature

The strongest objection to this whole comparison is that neither vendor publishes pricing, so any general verdict is moot until a firm runs a request for proposals and gets a real quote. That objection is fair. The smaller company's October round also priced below the $10 billion reported in August, a reminder that a reported target is not a valuation.

But the operational gap does not depend on what any round closes at. The embedded engineer model and the 35-hour onboarding are structural product decisions made well before this funding cycle. They held through both autumn rounds, September's and October's. Evaluate the deployment model. Not the term sheet.

Switching Cost: What Moving Actually Takes

The cost most comparisons skip is the exit. A firm that signs a multi-year Harvey contract with an embedded engineer has sunk six to nine months of that person's time into firm-specific workflows that do not transfer to a competitor. A firm on the lighter onboarding has less sunk cost, but also less firm-specific customization to lose.

Both companies also compete against an incumbent: Thomson Reuters' CoCounsel, which reported one million professionals across 107 countries and territories in a February 2026 release. That counts individual users inside firms that most likely already pay for the incumbent's research tools, not net-new logos. A firm already on those products has a lower-friction fallback if either startup's pricing moves against it later.

Before switching, check three things:

• Data export: whether custom agents, prompts and review tables can leave in a usable format.
• Retention: that the old vendor's deletion of your documents is confirmed in writing.
• Seat terms: the renewal uplift cap, which Vaquill reports runs 10 to 25 percent a year when nobody negotiates it.

The funding post that is the source for the closed valuation, captured 12 October 2026.

Sponsor spot open

What Would Change This Answer

Watch two numbers. First, the embedded engineer capacity: the outside estimate is 40 to 80 engineers, and a $550 million round buys more of them, but if the embed queue backs up, firms waiting for their engineer will feel it long before anyone reads a term sheet.

Second, what Legora does with its Series E. Its release says the money funds US and international expansion and its agent platform; if that buys an embedded engineering program, the deployment gap narrows, and if it does not, expect it to lean harder on the fast-onboarding pitch. Whichever one keeps its deployment promise as its valuation climbs is the one worth re-checking in April 2027.

Frequently asked questions

Is Harvey or Legora cheaper?

Legora is cheaper at the contract floor. Law XY AI reports a list price near $3,000 per user per year with a 10-seat minimum, about $30,000 a year, while analysts cited by Vaquill put Harvey's floor near $360,000 on 25 seats for 12 months. Neither company publishes prices, so treat both as estimates until a quote arrives.

How long does it take to onboard Harvey versus Legora?

Harvey typically places a Forward Deployed Engineer with a client full-time for six to nine months to build firm-specific workflows, according to Perspective AI's analysis. Legora's GÖRG case study describes 35 hours of face-to-face onboarding and 96.81% weekly active usage at a firm of more than 380 lawyers.

Which company has raised more funding, Harvey or Legora?

Harvey has raised more. Its March 2026 post put total funding above $1 billion, before a September 2026 round of $550 million at a $15.5 billion valuation. Legora raised $600 million across its two 2026 Series D tranches, then $550 million in a Series E on 21 October at an $8.8 billion valuation.

Which AI models power Harvey and Legora?

Harvey started on OpenAI's GPT-4 series models, and OpenAI took part in its $80 million Series B. In August 2026 it released Harvey Tenet, a research model post-trained from Moonshot AI's open-weight Kimi K3. Legora does not name its model providers on its security page, and says customer data never trains any model.

Is there a cheaper alternative to Harvey and Legora?

Thomson Reuters' CoCounsel is the established incumbent, which said in February 2026 that one million professionals in 107 countries and territories had chosen it. Firms already paying for Westlaw may find it the lower-friction option. It does not offer the same custom-agent depth as Harvey or the collaborative review tools Legora sells.

Covered in this guide

• Harvey: Legal AI platform serving 60 of AmLaw 100 firms. Document review 80x faster with citations, contracts, due diligence, and workflow agents. Enterprise-only.
• Legora: AI workspace for lawyers built on Claude, used for bulk Tabular Review, contract drafting, and agentic legal workflows at the enterprise tier.
• Anthropic: Anthropic is a Public Benefit Corporation started by 7 ex-OpenAI researchers. It builds the Claude model family and counts Amazon and Google among its largest backers.
• ChatGPT: ChatGPT is OpenAI's AI assistant, with 1.2 billion weekly users by OpenAI's count, the GPT-6 model family and plans that run from a free tier to a premium Pro tier.
• Claude: Claude is Anthropic's AI assistant, with a free tier, Pro, Max and Ultra personal plans, and Team and Enterprise seats. Opus access starts at Pro.
• GLM-5.2: GLM-5.2 is Z.ai's June 2026 open-source 744B MoE model with a 1M-token context window, 85.6% GPQA Diamond (Vals AI), and MIT license at $1.40/1M input tokens.
• Kimi K3: 2.8T-parameter open-weight MoE model from Moonshot AI (July 2026) with a 1M-token context window and the top published open-weight result on GPQA Diamond.
• Microsoft: Microsoft, founded 1975, is a public company (NASDAQ: MSFT) and the Redmond maker of Windows, Azure, GitHub and the Copilot assistants, with $331.8B of fiscal 2026 revenue.
• Moonshot AI: Beijing AI lab founded in 2023 by Yang Zhilin, maker of Kimi and the open-weight Kimi K3 models, valued at $20B at its May 2026 round.
• Nvidia's: Santa Clara chipmaker that supplies the GPUs, networking, CUDA software and open Nemotron models behind most large AI training and inference clusters.
• OpenAI: OpenAI builds ChatGPT, the GPT-6 family (Astra, Sol, Terra, Luna), Codex and the OpenAI API. It reports 1.2B weekly ChatGPT users and raised new funding in March 2026.
• RealPact: AI contract platform for real estate brokerages, backed by YC S26. Voice agents pull deed, tax, and MLS records to auto-fill contracts in minutes.
• Thomson Reuters' CoCounsel: Toronto-based $41.1B public company (est. 2008) delivering agentic AI solutions and trusted intelligence for 50K+ professionals globally.
• Z.ai: Z.ai (formerly Zhipu AI) is a Beijing lab behind the GLM model family, founded 2019 and listed on the Hong Kong Stock Exchange (2513) since January 2026.

Sources