At Inscribe, Ivo’s first user conference, we announced three developments in our mission to push contract intelligence forward: Ivo Collaborate, a new way to leverage and apply contract intelligence across intake, negotiation, and approval; the Ivo–Micro1 Contract Bench, a benchmark for measuring AI’s actual judgment and effectiveness in contract review; and Ivo Sage, an open-source AI model post-trained for long-horizon contract work.
Together, these announcements reflect a broader shift away from software that simply organizes contracts workflows towards technology that reflects the way lawyers actually work while enabling them to uplevel their value to the business. With these releases, Ivo aims to not only reveal the business intelligence hidden within contracts, but to provide the frameworks and technology that will advance it even further.
Introducing Ivo Collaborate
Ivo Collaborate brings the entire contracting process into one place, from intake through negotiation, approval, and signature. Collaborate gives the negotiation process itself a system to live in: it allows the decisions, positions, and concessions that shape each deal to be managed in context and in one location, all while retaining the data from that process as intelligence that your team can utilize going forward.
Unlike how a CLM simply routes documents based on a predefined field, Collaborate actually reads each contract as it arrives, then assesses the agreement against the playbooks and rules set by the legal team before routing it accordingly. Low-risk agreements can move directly to signature, while those with risks that require further judgment are escalated to the right reviewer, with all the relevant context already identified.
With Collaborate, anyone in the organization can start a deal request directly, removing the traditional legal bottleneck at intake while making sure that legal’s judgment is consistently and effectively applied across all the organization’s agreements. The result is a system where legal can define how the business contracts across every deal without the need for laborious manual review, and all while capturing what the organization learns along the way—turning every negotiation into an opportunity to make the next deal smarter.
Read more about Collaborate on our blog here.
A benchmark for measuring the judgment behind contract review
Ivo has released a new contract review benchmark, developed in collaboration with data lab and research partner Micro1. The Ivo–Micro1 Contract Bench was developed to evaluate aspects of contract review that conventional AI benchmarks can miss. Many existing approaches to evaluating AI models for contract work focus on whether the model can simply identify issues and make edits. However, real contract review requires knowing things like what not to change, how to adapt a position to the deal’s circumstances, how to execute established positions in context, and how to effectively recognize when an issue should be escalated to a human. Most models aren’t judged on these kinds of nuances, like the precision, efficiency, and effectiveness of edits.
The Ivo–Micro1 Contract Bench evaluates contract review effectiveness based on the decisions and judgement calls that real attorneys make. Our contract bench evaluates models across five dimensions: issue prioritization, negotiation restraint, situational judgment, escalation judgment, and standard-position execution. Ivo–Micro1 Contract Bench thus uses contract-review tasks, detailed playbooks, deal context, and counterparty redlines in its review, with the criteria used to evaluate model decisions established by real attorneys.
Early findings show that, against these benchmarks, leading models concede easily and push back poorly: when the right response is to counter or reject a provision, they get it right only 46% of the time. When the right response is to escalate an issue to management, the leading models meet only 23% of the escalation criteria. These results highlight a real discrepancy between existing review standards and criteria that would lead to meaningful and effective review.
A public set of these benchmark results, along with the dataset used for the benchmark criteria, will be available in the coming weeks.
Introducing Ivo Sage
Ivo research efforts have also produced Ivo Sage, an open-source model that is the first to be post-trained for long-horizon contract work. General-purpose frontier models are built to handle a wide range of tasks; however, contracting requires a much more specialized set of skills. Contracts are long and complicated documents, necessitating multi-step legal reasoning and deep context to adequately review. Post-training gives an AI model the opportunity to learn these patterns and tasks more effectively and surgically, rather than approaching every contract as a general-purpose problem.
The Ivo Sage model was built in partnership with River AI by post-training DeepSeek V4 Flash specifically for complex contract work. This was done 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 70% of the pass criteria of the Legal Agent Benchmark (LAB) Contracts to 91%. Overall, Ivo Sage reached comparable quality of much larger frontier models, with higher token efficiency at a fraction of the cost.
Unlike other legal AI companies, Ivo is releasing the Ivo Sage mode as an open-source model. Ivo Sage is being released along with its research and evaluation record so that researchers, developers, and legal tech builders can download it, study how it performs, and build on it. Our goal is to make progress in legal AI more transparent and collaborative. By sharing our model and the work behind it, Ivo is hoping to give others a foundation they can test and improve as contract-specific AI continues to develop.
Ivo Sage is available to download on Hugging Face here.
Building what comes next
Ivo is committed to making everything surrounding contracts better, smarter, and more intelligent. That means building technology that can meet the complexity of legal work, creating new ways to improve what AI can do, and making the underlying research accessible enough to move the whole field forward.
Our work is still evolving, but the direction is clear: a future where contracts are not just documents to manage, but intelligence that moves the business.
