Researched
This analysis is based on documentation, public user reports, and vendor materials — not yet on our own hands-on testing. How we rate
The most expensive way to be right about legal AI
Harvey is the best-funded company in legal AI by a wide margin: roughly $11 billion valuation as of March 2026, over a billion dollars raised, and reported ARR near $190 million that nearly doubled in five months.
The product is good. Over 100,000 lawyers across 1,300-plus organizations use it, including a majority of the Am Law 100. For research, litigation support, transactional work and contract analysis across practice groups, it does what it says.
The question worth spending this page on isn't whether Harvey works. It's whether you're the customer it's priced for, and the answer for most firms reading a page like this is no.
The pricing inverts what you'd expect
Harvey publishes nothing. Every number here is reported by industry press, procurement analysts or customers, and should be treated as a planning range rather than a quote.
| Firm size | Reported per-seat monthly |
|---|
| Mid-market (~50–200 attorneys) | ~$1,000–$2,000 |
| Am Law 100 scale | ~$100–$200 |
Read that twice, because it's the most important thing on this page. Smaller firms reportedly pay up to ten times more per seat than the largest ones.
Volume discounting reportedly kicks in above roughly 100 seats. Below that, you're at the top of the ladder. A 60-lawyer firm and a 600-lawyer firm buy the same software at radically different unit prices, and the structure is optimized for the buyers who need the break least.
Annual contracts commonly run $50,000 to $300,000-plus, with reported seat minimums of 25 to 50 on 12-month terms. A frequently cited floor: around $288,000 a year for a 20-seat deployment at roughly $1,200 per seat.
There's also a reported tier structure where a base seat sits near $1,200 monthly and a LexisNexis-bundled seat closer to $2,400. Bespoke custom-model builds exist at the top end — Harvey's CEO has publicly said such a build could exceed $5 million.
The renewal is where it gets you
This is the item procurement should care about most, and it's the one buyers consistently underweight.
Reported customer experience through 2025 and 2026 describes annual renewal uplifts of 10 to 25 percent absent a contractual cap. Analysts suggest pushing for 5 to 7 percent as a reasonable ask.
There's a subtler version too. The reported "price increase" isn't always a list price jump — it's the bundle spreading. Firms that started on a pilot rate or a lower tier find renewal lands higher up the ladder as add-ons attach themselves to the contract.
Given how fast Harvey's valuation has moved — $5 billion mid-2025, $8 billion in December, $11 billion in March — there's obvious upward pressure on per-seat economics. Four-figure annual contract values are how those multiples get justified.
Put the cap in writing at signature. Aggressively. It's the clause your successor will be grateful for.
What you're actually buying
Beyond the model, which Harvey doesn't own, the value is scaffolding: legal-specific safeguards, document management integrations, citation grounding into the LexisNexis corpus, agentic workflows that complete multi-step tasks, and white-glove onboarding.
That scaffolding is real and it's most of the product. The LexisNexis partnership in particular gives access to a research corpus competitors can't replicate, which matters for firms doing work across obscure jurisdictions.
What it isn't: a fundamentally different reasoning engine. Harvey reportedly runs partly on frontier models available directly from the labs that make them. That's not a criticism — assembling the legal layer around a good model is the hard part — but it does inform the small-firm calculation below.
How we researched this
We're not a Harvey customer, and on a product sold through six-month enterprise cycles, essentially nobody writing publicly is.
This page triangulates industry reporting, procurement analyses and customer disclosures from 2025 and 2026. Funding and valuation figures come from mainstream coverage — CNBC on the March 2026 round, Yahoo Finance on December 2025, TechCrunch on ARR — and are the most solid facts here. The pricing figures are the least solid, and we've labeled them as reported throughout.
A source problem to name plainly: nearly every detailed Harvey pricing analysis available publicly is published by a competitor — contract lifecycle vendors, cheaper legal AI tools, drafting products, and at least one site built entirely around recommending a general assistant instead. Their reported figures are broadly consistent with each other and with the industry press, which is why we've used them. Their conclusions are sales arguments and we've set those aside. The specific claim that small firms should use a general assistant instead comes from a source with an obvious interest in it — we think the underlying logic holds for solos, and you should know where it originated.
Estimates in circulation vary wildly, from $500 per seat per year at one end to $2,400 per seat per month at the other. That spread tells you these are triangulations, not disclosures.
Where this doesn't make sense
Firms under about fifteen lawyers. You'd pay top-of-ladder rates for M&A agents you'll never open. Over 80% of US law firms have five or fewer attorneys, and for them $1,200 a seat isn't a pricing question — it's a non-starter competing against rent and malpractice insurance.
Solo practitioners. No procurement committee, no IT team, no budget line. And a general assistant is already open in another tab at $20 a month.
Anyone needing an API. Harvey is a platform, not developer infrastructure.
Firms with uneven usage. Unlimited seats reward daily use. If your AI-heavy work concentrates in a few matters annually, ask about the reported credit-metered option rather than accepting per-seat.
Contract management specifically. If drafting, reviewing, negotiating and managing contracts is the actual problem rather than broad legal work, dedicated CLM tools solve it at published, far lower prices.
Against the alternatives
Against Legora: the closest peer, also quote-only, reportedly around 800 customers across 50-plus markets with a $5.55 billion valuation. Broadly: Harvey for US-headquartered firms wanting US coverage and brand recognition; Legora for European operations where data residency and European legal coverage matter. Worth pricing both — a credible alternative in the room is the main lever on an unpublished price.
Against CoCounsel: bundled with Westlaw, published pricing in the low hundreds monthly. Narrower, dramatically cheaper, and sufficient for a lot of firms.
Against Claude or ChatGPT: the same frontier reasoning at roughly $20 to $30 per user per month, no seat minimum, month-to-month. You lose the legal safeguards, the licensed research corpus, the DMS integrations and the workflow scaffolding. For a solo drafting and summarizing, that trade is usually right. For a firm running complex transactional workflows, it usually isn't — and confidentiality and privilege considerations need their own analysis before you paste anything client-related anywhere.
Against Hebbia: overlapping on deep document analysis and used by law firms as well as funds. Reported seat pricing is dramatically lower. Different architecture, worth comparing directly if document interrogation is the core need.
Against Glean: enterprise search across your own systems rather than legal-specific reasoning. Complementary more than competitive.
Pricing 2026
| Item | Reported figure | Confidence |
|---|
| Mid-market seat | ~$1,000–$2,000/user/mo | Reported, widely consistent |
| Am Law 100 seat | ~$100–$200/user/mo | Reported |
| LexisNexis-bundled seat | ~$2,400/user/mo | Reported |
| Annual contract | $50,000–$300,000+ | Reported |
| Seat minimum | 25–50, 12-month term | Reported |
| Renewal uplift | 10–25%/yr uncapped | Reported customer experience |
| Custom model build | Can exceed $5M | Stated publicly by Harvey's CEO |
| Credit-metered plan | Exists, terms unknown | Reported, unconfirmed |
Checked August 2026. Harvey publishes no pricing; every figure above is triangulated from industry reporting, procurement analyses and customer disclosures, much of it published by competitors. Public estimates in circulation range from $500 per seat annually to $2,400 per seat monthly. Funding: $200M at a reported ~$11B valuation in March 2026, up from $8B in December 2025. Get a written quote — you cannot budget for Harvey from any web page, this one included.
Cap the renewal at signature. Ten to twenty-five percent uncapped uplift compounds fast. Five to seven percent is the reasonable ask.
Itemize the quote. The LexisNexis bundle roughly doubles the reported seat rate. Decide whether you need the corpus before accepting it.
Price Legora in parallel. A credible alternative is the only real leverage on an unpublished price.
Count your seats honestly. Minimums of 25 to 50 mean small firms buy licences nobody opens.
Our Verdict
Harvey is genuinely excellent enterprise legal AI, deeply integrated, and built for the largest legal organizations in the world. For firms with the document volume, matter complexity and IT budget to absorb a six-figure commitment, it earns its position and the lack of public pricing is normal for its market rather than a warning sign.
The pricing structure is the story. Reported gaps of up to ten to one between mid-market and Am Law seat rates mean smaller firms subsidize the largest ones. That's a defensible commercial strategy and it makes Harvey structurally wrong for exactly the firms most likely to be reading about it.
Renewal terms deserve more attention than the headline rate. Uncapped uplifts of 10 to 25 percent, plus the bundle-creep effect, mean year three costs meaningfully more than year one unless you negotiated for it.
The valuation creates pressure. A company repriced from $5 billion to $11 billion in nine months has investors expecting the revenue curve to keep pace, and four-figure seat values are how that happens.
For an Am Law firm or a large in-house department with a procurement function: worth a serious evaluation, and the volume discounts make the economics work. For a mid-size firm: run the math carefully against the feature set you'll actually use, because you're paying the highest rate on the ladder. For solos and small practices: this isn't a pricing objection you can negotiate around — it's the wrong product, and a general assistant plus a focused drafting tool will serve you better for a rounding error of the cost.
Note: AIVario earns no commission from Harvey. This page is based on published reporting, procurement analyses and industry sources rather than a deployment. We are not a law firm and this is not legal technology advice — evaluate confidentiality and privilege implications with your own counsel.
Best for: Am Law 100 firms, large in-house legal departments, firms running complex transactional and litigation workflows at volume, organizations with procurement and IT capacity
Not ideal for: Solo practitioners, firms under fifteen lawyers, mid-size firms without heavy AI workflows, teams needing an API, anyone requiring published pricing
Bottom line: The category leader, priced so that the firms who need discounts least receive them. Excellent if you're Am Law scale; structurally the wrong purchase if you're not.
- Hebbia — deep document analysis used by law firms and funds, reportedly far cheaper per seat
- Claude — the underlying reasoning without the legal scaffolding, at 1/40th the cost
- Glean — enterprise search across your own systems
- AlphaSense — the equivalent platform on the financial research side
- ChatGPT — general assistant many small firms use instead
Frequently Asked Questions about Harvey
What does Harvey cost?
Harvey publishes nothing, so every figure is reported rather than official. Industry reporting and customer disclosures through 2026 put mid-market firms of roughly 50 to 200 attorneys at approximately $1,000 to $2,000 per user per month, falling to roughly $100 to $200 per seat at Am Law 100 scale. Annual contracts commonly land between $50,000 and $300,000-plus, with reported seat minimums of 25 to 50 on 12-month terms. There's no free tier, no monthly plan and no self-serve signup. A commonly cited floor is around $288,000 a year for a 20-seat deployment at roughly $1,200 a seat.
Why do small firms pay more per seat than big ones?
Volume discounting, which reportedly starts to bite above roughly 100 seats. The practical result is that a 60-lawyer firm and a 600-lawyer firm buy the same product at radically different unit prices — reported gaps run as high as ten to one. That means the pricing is structurally optimized for the firms with the least need for a discount. If you're a 40-attorney practice, you're paying the highest rate on the ladder for a feature set built around workflows Am Law firms run and you probably won't.
Should I worry about renewal increases?
Yes, and this is the line item to negotiate hardest. Reported customer experience through 2025–2026 describes annual renewal uplifts of 10 to 25 percent without a contractual cap. Procurement analysts suggest pushing for a 5 to 7 percent cap as reasonable. There's also a compounding effect that isn't a simple price rise: firms that started on a pilot rate or a lower tier tend to find renewals land higher up the ladder as add-ons get attached. Get the cap in writing at signature — it's far easier than fighting it later.
What's the LexisNexis element?
Harvey has a data partnership with LexisNexis giving it access to a proprietary legal research corpus that smaller competitors can't match. For firms doing deep research across obscure jurisdictions, that's a genuine advantage. It also costs: reported seat tiers describe a base seat near $1,200 per month and a LexisNexis-bundled seat closer to $2,400, and the partnership is expected to push all-in costs meaningfully higher for firms wanting the bundled content. Decide whether you need the corpus before assuming the bundle.
How big is Harvey really?
Very. The company raised $200 million in March 2026 at a reported $11 billion valuation, co-led by GIC and Sequoia, up from $8 billion three months earlier and $5 billion in mid-2025. Reported ARR reached roughly $190 million by early 2026, nearly doubling from $100 million in about five months. It reports over 100,000 lawyers across 1,300-plus organizations in 60-plus countries, including a majority of the Am Law 100, with firms like A&O Shearman and Latham & Watkins publicly associated. Total funding exceeds $1 billion.
Can a small firm get most of this cheaper?
A meaningful portion, yes — with an important caveat. Harvey reportedly runs partly on frontier models from major AI labs, and using one of those assistants directly gives you the same underlying legal-reasoning engine for drafting, summarizing and contract analysis at a fraction of the cost, with no seat minimum. What you don't get is the legal-specific safeguards, document management integrations, citation grounding into a licensed research corpus, and the workflow scaffolding that constitutes most of Harvey's value. For a solo practitioner the trade often makes sense. For a firm handling complex transactional work at volume, it usually doesn't.
How long does buying it take?
Sales cycles are reported at six months or longer, and implementation requires enterprise procurement, IT involvement and often a dedicated onboarding team. This is not a tool you evaluate over a weekend. For Am Law 100 firms with technology budgets in the millions and staff to manage a rollout, that's normal. For everyone else it's a substantial hidden cost in internal time on top of the contract value, and it should factor into the decision alongside the licence fee.
Is there a usage-based option?
Reported coverage describes a pay-as-you-go, credit-metered plan alongside the per-seat model, which would suit firms with uneven usage better than unlimited seats. Details aren't public and we can't confirm the terms. If your usage is concentrated in a few matters per year rather than spread daily across a practice group, it's worth asking about specifically — it's the kind of option that doesn't get offered unless you raise it.
