Introducing Tenet, our first model post-trained for legal.
Tenet is a Kimi K3 base that we post-trained with @FireworksAI_HQ on a corpus of publicly available legal data, synthetic data, and human expert data simulating long-horizon legal work.
Training increases Tenet's all-pass rate by 82% on LAB and 22% on LAB Contracts relative to the Kimi K3 base model. It achieves state-of-the-art performance on LAB Contracts and places second on LAB.
These gains generalize to other leading agentic benchmarks including @mercor's Apex Agents - Corporate Law, @crosbylegal's Redline Bench, and @scale_AI's Professional Reasoning Bench.
Tenet is also optimized for token efficiency, operating at less than a fourth the cost of leading foundation models.
We additionally post-trained three specialist models for Tenet to use as subagents:
1) M&A Diligence: post-trained with @baseten on our LAB Diligence environment in an RLM harness, this model is optimized for high-scale, long-horizon tasks.
2) Review Tables: trained with @appliedcompute on our Review Table environment, this model is state-of-the-art and cost-effective at high-volume document review and structured data extraction.
3) Firm Knowledge: trained with @EngramLab on our synthetic law firm environment, this model is optimized to learn and search over a firm's knowledge via memory and structured notes.
More details on model training, environment design, benchmarking, results, and more in the article by @gabepereyra below.
What's next for Harvey’s research?
- Scaling LAB to more jurisdictions, practice areas and workflows
- Scaling compute to bring new generalist models and capabilities to Harvey
More to come soon.
Here’s what we’re building at Harvey to make verification faster:
• Check quotes and pin cites against full opinions
• See what agents read, searched for, or couldn’t open
• Track who reviewed what and who signed off
Harvey’s approach to verification: https://t.co/u7iqFyGBOh
A lawyer signs their name to every piece of work that leaves their desk, even when Harvey wrote the first draft.
As Harvey reviews tens of thousands of documents in one table or works through an entire data room for a diligence memo, human verification can become a bottleneck.
Patent research often means following a claim across scattered records: the issued patent, prosecution history, and later proceedings. A larger index alone doesn’t tell an agent which record to check next.
We built Harvey’s patent research agent to make those decisions as it works. It identifies what the answer needs to establish, chooses sources that can supply the evidence, and uses each result to decide what to investigate next.
That feedback loop is what the diagram shows: each discovery can reshape the research path toward a cited answer—or reveal a gap that remains unresolved.
See how we built it: https://t.co/M6FPDRwVAO
Two courts, one standard.
Harvey is the new Official Legal Assistant Partner of @Wimbledon . See you at the The Championships in 2027.
Learn more: https://t.co/avacbLlsy6
How Phil built a collaborative Word editor with no human engineers.
Inspired by @russelljkaplan's story of how @cognition builds their Android app with just a swarm of Devins. https://t.co/6dDxnygL0m
Coming soon to Harvey: GPT-6.2 Astra from OpenAI.
In our early testing, GPT-6.2 Astra more than tripled GPT-6 Astra’s all-pass rate on LAB, with strong results in international trade and sanctions, environmental and ESG, and litigation. It also showed strong judgment on contested matters, flagging a client’s weak spots for the legal team in the internal memo while keeping them out of the filing, where they could help the other side.
GPT-6.2 Sol is also coming soon, with a higher all-pass rate than GPT-6.1 Sol across most practice areas at less than half the output tokens.
Learn more: https://t.co/RF1X5TZ8fg
Legal help, right in Slack.
Harvey Intake gathers context, drafts answers using your team’s policies, and routes them for legal sign-off before replying in the same thread.
Faster answers for the business, with Legal in control.
Learn more: https://t.co/MVp6srbZ0y
Harvey’s platform has completed an independent IRAP assessment at the PROTECTED level.
The assessment examined applicable controls from the Australian Government Information Security Manual, giving public sector organisations independent evidence to inform their own risk assessments and authorisation decisions.
We’ve also been approved as a supplier under the NSW Government ICT Services Scheme and are listed on Victoria’s Digital Marketplace for eServices, giving eligible buyers established ways to procure Harvey.
Australia is a priority region for us, and we’re committed to long-term partnerships with public sector legal teams.
Read more: https://t.co/aiQm1q77CF
Harvey was named to the inaugural 2026 AI 100 from @NewcomerMedia and @BessemerVP, recognizing the top 100 private AI companies in the world.
Thank you to our customers who push us to make Harvey better, and to our team who makes it happen.
See the full list: https://t.co/epiLiV9CXV
Bringing ethical walls into legal AI means protecting confidential information wherever it’s used.
Our engineering team explains how they built Harvey to enforce firms’ ethical walls across documents, conversations, and shared work.
Read more: https://t.co/t1QZUmuMys
The latest redline rarely tells the whole story.
Contract Intelligence now keeps the full deal history with the contract.
Review each new round with prior versions and related agreements in view.
Track every contract through to signature. Let signed agreements inform your next review.
Learn more: https://t.co/6y9Suu2vem
Six Harvey Legal Engineers, all former lawyers, share how they’d use Harvey Agents for Outlook to prioritize emails, manage calendars, kick off legal work, and organize matter correspondence, spending less time managing their inbox and more time on substantive legal work.
Danish legislation is coming to Harvey.
Through our new partnership with Djøf Forlag, lawyers with a Jurabibliotek license will be able to research and draft with Danish acts, executive orders, and preparatory works. They can do this alongside their own documents and prior advice, with every answer citing its source.
Learn more: https://t.co/ocrsRkXgTE
Coming soon to Harvey: Gemini 4 Carbon, @GeminiApp's latest model.
In our early testing, Gemini 4 Carbon was especially strong on deal management and case law research, and its accuracy held up on long, document-heavy work.
What can you expect from Harvey in 2027?
Anique Drumright, Harvey’s Chief Product Officer, shares the vision for a product that’s more connected, more specialized, and more personalized to the way you work.
#FORUMChicago
We're expanding our partnership with @Wolters_Kluwer to bring CCH iKnowConnect content into Harvey.
Building on the Wolters Kluwer primary law already in Harvey, CCH iKnowConnect subscribers in Australia, New Zealand, Singapore and Malaysia will be able to access editorial commentary, case headnotes and practical guidance directly in Harvey, with citations linking back to the source.
With offices in Bengaluru, Sydney, Singapore, and Tokyo planned for early 2027, this partnership deepens Harvey's commitment to legal and tax professionals across APAC.
Read more:
Coming soon to Harvey: Fable 5.2 from Anthropic.
In early testing, Fable 5.2 showed gains in immigration work, giving clearer recommendations on client decisions and next steps than Fable 5.1.
In an I-9 review, for example, it flagged urgent compliance issues and gave employee-specific instructions for correcting them.