# Harvey agents by Harvey Canonical page: https://technooptimist.io/ai/harvey-agents Frontier: [AI](https://technooptimist.io/ai.md) · Field: Agents · Company: [Harvey](https://technooptimist.io/companies/harvey.md) Stage: Scaling (5 of 5: Research → Proto → Pilot → Deploy → Scaling) Updated: last update 2026-09-09, facts checked 2026-10-11 Goal: Let AI do the bulk of legal grunt work, reading data rooms, drafting memos and marking up contracts, so lawyers review and decide instead of spending days on documents. Status: Used by most top US law firms; own model Tenet in research preview since Aug 2026. Next milestone: Tenet deployed inside Harvey's product ## Milestones | Status | Milestone | Date | Slip | | --- | --- | --- | --- | | Done | Passes $100M annual recurring revenue (estimate) | Aug 2025 | | | Done | $200M at $11B valuation to expand agents | 25 Mar 2026 | | | Done | Legal Agent Benchmark released open source | 6 May 2026 | | | Done | Tenet, first in-house model, in research preview | 20 Aug 2026 | | | Done | $550M at $15.5B valuation | 9 Sep 2026 | | | Current | Tenet deployed inside Harvey's product | | | ## Most important updates - **9 Sep 2026**: Harvey raises $550M at a $15.5B valuation. Harvey raised $550 million co-led by Diffusion and Lightspeed, with Sequoia, Kleiner Perkins and a16z returning and Sapphire and Whale Rock joining. ([our story](https://technooptimist.io/stories/harvey-raises-550m-at-a-15-5b-valuation-b183caef.md) · [original source](https://www.harvey.ai/blog/harvey-raises-dollar550m-at-a-dollar155b-valuation-to-help-legal-teams-own-their-intelligence)) - **20 Aug 2026**: Harvey previews Tenet, its first own legal model, post-trained from Kimi K3. Harvey and Fireworks AI used reinforcement learning on long legal tasks to post-train Moonshot's open-weight Kimi K3 into Tenet. ([our story](https://technooptimist.io/stories/harvey-previews-tenet-its-first-own-legal-model-post-trained-from-kimi-6ff4a238.md) · [original source](https://fireworks.ai/blog/post-training-kimi-k3-with-harvey-for-long-horizon-legal-work)) - **6 May 2026**: Harvey open-sources the Legal Agent Benchmark: 1,200+ tasks graded by lawyers. Harvey released LAB, an open benchmark of more than 1,200 agent tasks across 24 practice areas, graded by over 75,000 expert-written criteria. ([our story](https://technooptimist.io/stories/harvey-open-sources-the-legal-agent-benchmark-1-200-tasks-graded-by-0884e566.md) · [original source](https://www.artificiallawyer.com/2026/05/06/harvey-launches-legal-agent-bench/)) - **25 Mar 2026**: Harvey raises $200M at an $11B valuation to expand its legal AI agents. Harvey raised $200 million co-led by existing investors GIC and Sequoia, valuing it at $11 billion, up from $8 billion three months earlier. ([our story](https://technooptimist.io/stories/harvey-raises-200m-at-an-11b-valuation-to-expand-its-legal-ai-agents-ccd45cc0.md) · [original source](https://news.bloomberglaw.com/bloomberg-government-news/legal-ai-startup-harvey-raises-funds-at-11-billion-valuation)) ## Current obstacles - **All-or-nothing quality bar**: A legal memo that misses one issue is not usable; top models still fully pass only about 1 in 5 LAB tasks. - **Licence terms on Tenet's base**: Kimi K3's licence needs a separate deal with Moonshot for services earning over $20M a year, which Harvey exceeds. ## Physics limits - **Models can invent citations**: Language models produce fluent text that can cite cases or clauses that don't exist, so every point needs checking against sources. - **Long tasks compound errors**: Over 50+ steps and 100,000 generated words, small per-step error rates multiply into a high chance of at least one mistake. ## How it works ### Matter: Load the client's documents A lawyer gives a request and the matter's documents, such as a data room for an acquisition, in a secure vault. ### Plan: Split the job The agent breaks the request into steps: find relevant clauses, check each against the question, note risks. ### Execute: Read, search, draft It reads and cross-checks hundreds of documents with tools and models, then drafts the memo or markup. ### Review: Lawyer signs off Output cites its sources so a lawyer can verify each point before it goes to a client. ## Spec sheet | Spec | Harvey agents | Provenance | | --- | --- | --- | | Am Law 100 firms using Harvey | 80% | reported | | Tenet LAB all-pass rate | 19.7% | reported | | LAB tasks / practice areas | 1,200+ / 24 | reported | | Tenet cost per LAB task | $5.92 | reported | | Annual recurring revenue | $400M+ (Aug 2026, estimate) | estimated | Provenance: reported = stated by the company; estimated = our estimate; sample = a sample figure. ## Papers and demos - **6 May 2026** (dataset): [Legal Agent Benchmark (LAB): 1,200+ legal agent tasks, open source](https://www.artificiallawyer.com/2026/05/06/harvey-launches-legal-agent-bench/). Gives the field a shared, lawyer-graded test of agents doing real legal work. ## About Harvey Harvey sells AI agents to law firms and corporate legal teams for research, drafting, contract review and due diligence. The largest legal AI company, valued at $15.5B in Sept 2026. - Used by 80% of the Am Law 100 and five of the Fortune 10, per Harvey. - Tenet (Aug 2026), its first own model, post-trains Kimi K3 for long legal tasks and nearly doubles completion on LAB. - Valuation rose from $8B (Dec 2025) to $11B (Mar 2026) to $15.5B (Sept 2026). - Founded: 2022 - Headquarters: US - Status: private - Valuation: $16B (private valuation, 2026-09) - Raised: $750M - Last round: Funding round $550M, 9 Sep 2026, led by Diffusion, Lightspeed Venture Partners - Website: https://www.harvey.ai - People: Winston Weinberg (Co-founder & CEO); Gabe Pereyra (Co-founder) - Partners, customers, investors: Fireworks AI (supplier); Moonshot AI (supplier); Sequoia (investor) Company page: https://technooptimist.io/companies/harvey.md ## Related programs Everything in AI: https://technooptimist.io/ai.md --- The Techno Optimist tracks every frontier of technology, program by program: https://technooptimist.io/ (a Markdown version of any page: add .md to its URL; site map for agents: https://technooptimist.io/llms.txt). Data API (JSON, realtime stream): https://technooptimist.io/developers. People read 5 pages free, then Reader $10/month; these Markdown pages are open to agents.