Ivo发布首个面向长期合同工作的开源法律AI模型

Unlike other legal AI companies, Ivo Sage is the first model that attorneys, researchers and developers can download, run and build on top of it for free; Ivo also previews a new contract-review benchmark at Ivo Inscribe, the company’s first customer event for in-house legal and contracting teams SAN FRANCISCO, CA – Inscribe Conference – Oct. 1, 2026 — Ivo , the leading contract intelligence platform for major global enterprises, today became the first legal AI company to release an open-source model post-trained for long-horizon contract work. Announced at Ivo Inscribe , the company’s first user conference, Ivo Sage gives developers, researchers and legal teams access to AI technology they can test, train and build on. Legal work looks different from one organization to the next, from the complexity of a Fortune 500 company to a small business that handles a more narrow set of agreements. By open-sourcing its model, Ivo is enabling legal practitioners, AI researchers and developers to fine-tune it on their own datasets, adjust its rules and safeguards for specific use cases, and iterate quickly on new ideas or approaches. Ivo is sharing its work to drive more transparent innovation in the legal field, with every improvement open for the broader community to test, verify, and build upon. “The next leap for AI in contract and legal work won’t come from bigger models. It will come from giving models the context of the work: the documents, the playbooks and the way a legal team operates,” said Min-Kyu Jung, co-founder and CEO of Ivo. “That’s the hard, unsolved part of legal AI, and it’s where we’re focused on next. We’re opening up our model so others can test it, adapt it and build on it. That’s how we move closer to contract intelligence the whole industry can rely on.” The Ivo Sage model was built in partnership with River AI by post-training DeepSeek V4 Flash on long-horizon contract work using both public and synthetic data generated by real attorneys, with River providing valuable technical and training infrastructure support. After reinforcement learning, the model went from scoring seventy percent of the pass criteria of the Legal Agent Benchmark (LAB) Contracts to ninety-one percent. It reached comparable quality of much larger frontier models and higher token efficiency at a fraction of the cost. Introducing a new benchmark: Ivo-micro1 Contract Bench While other legal benchmarks grade only the edits a model makes, Ivo’s new benchmark builds better judgement and escalation reasoning into the contract review process Contract review is a hard test for AI. Today, the technology is only graded by the edits a model makes, however, most of the judgement required happens before any edit. A good reviewer must know when to propose a change, when to leave acceptable language alone and when a decision needs to be escalated for an attorney’s judgement call. At Inscribe, Ivo previewed a new contract review benchmark in collaboration with data lab and research partner micro1 . The Ivo-micro1 Contract Bench is configured to measure those distinctions, across five dimensions of what a good attorney actually does: which issues to prioritize, when to show restraint, how to adapt to the specific deal, when to escalate, and how it follows your playbook. Early Findings: Frontier models can only take contract review so far. Strong judgement requires context. Ivo’s research found that leading models handle certain contract decisions well and miss others entirely, especially when to push back and when to escalate. Four early findings: Models are pushovers. They concede easily and push back poorly When the right response to a counterparty’s change is to accept it or leave it, models get it right 82 percent of the time. When it’s to counter or reject, they get it right 46 percent of the time, and they add a needed new point of their own only 23 percent of the time. Models tend to make their own calls instead of escalating to management Models don’t know when to ask. On average, models meet only 23 percent of the escalation criteria our expert attorneys set. That’s despite explicit instructions and a dedicated tool for when and how to escalate. This indicates models make edit decisions on their own rather than seeking permission and guidance from authority. GPT-6 Astra is the outlier at 71 percent. It gets there by escalating nearly everything, so it meets only 31 percent of its edit criteria and finishes last overall. Models make edits in broad strokes Attorneys make 64 percent of their changes inline, averaging 92 characters. On the same contracts, models make 14 percent to 37 percent of their changes inline, and their edits rewrite whole blocks of text at 207 to 369 characters. Every model does worse when the deal changes the answer On average, models meet 51 percent of the criteria for applying a playbook’s standard positions. When the right move depends on the facts of the deal instead, that falls to 38 percent. Every model drops by 9 to 15 points. The playbook often still supplies the rule; recognizing when the deal calls for it is where models fall short. A public set of the benchmark results will be available in the coming weeks. Ivo unveils a full end-to-end contract intelligence platform Ivo also announced the general availability of Ivo Collaborate, providing Fortune 500 legal teams with an end-to-end contract intelligence platform that runs the entire life of a contract, before and after signature. Collaborate reads incoming contracts and manages their entire workflow from intake through to signature, moving routine agreements according to rules set by legal and flagging issues that need a person’s judgment. Intelligence lets teams search signed agreements to understand what the business has promised and what it is owed, surfacing patterns within your portfolio that provide unprecedented business insights. Ivo’s broader vision is to make contracts a source of intelligence across the entire enterprise, serving as the core infrastructure layer that legal, procurement, finance, and sales teams rely on to review and negotiate agreements faster, maintain a single source of truth across their entire contract portfolio, and surface the business intelligence that informs better decisions. About Ivo Ivo is an AI-native contract intelligence platform built for the enterprise that helps legal and business teams move faster, reduce risk, and grow revenue by turning contracts into actionable business intelligence. Founded in New Zealand and headquartered in San Francisco, Ivo enables organizations to surface risk instantly, accelerate negotiation cycles, and remove bottlenecks across the contract lifecycle. Customers include leading global enterprises and high-growth companies across multiple industries. Media Contact LaunchSquad for Ivo ivo@launchsquad.com This content is labeled as created by a human - more information The post Ivo Becomes The First Legal AI Company to Publish a Free Open-Source Model Post-Trained for Long-Horizon Contract Work first appeared on The Legal Wire .