Building a Legal Coworker with GPT-5

Today, OpenAI announced the public release of GPT-5 – a novel frontier model family that combines reasoning and non-reasoning capabilities, with improved tool use, drafting, and user controllability.

During the early-access period, GPT-5 showed impressive performance across Harvey’s core product surfaces – Assistant, Vault, and Workflows – while also demonstrating substantial improvements in legal reasoning, as measured by expert preference and our BigLaw Bench evaluation suite.

We are thrilled to announce that GPT-5 is live on the Harvey platform via model selector today! We are actively incorporating GPT-5 as the default model where we see higher performance than other models. More information to come.

Last year, when OpenAI released their o1 series of reasoning models, it laid the groundwork for our mission to build collaborative agents that work alongside legal experts on their most complex work. Since then, we’ve furthered that vision with powerful new features like Workflows, Workflow Builder, and Deep Research. Together, these features are transforming Harvey from a copilot into a collaborative project partner.

But previous generations of models, while powerful, often forced a choice between long-form reasoning on a specific task or efficient processing of routine information. GPT-5’s novel architecture removes this distinction. Its ability to transition between in-depth reasoning and high-throughput, non-reasoning tasks fits seamlessly into Harvey’s flexible platform.

GPT‑5 lands at the perfect moment. Its stronger reasoning and tool orchestration will enable us to support more complex workflows that can branch, loop, and self-correct – all while keeping users in the driver’s seat through interaction and verification. Moreover, its long-form writing capabilities allow us to move beyond generating paragraphs to composing complete, multi-section documents that exhibit structural integrity and internal consistency, conforming to specific precedents and drafting conventions.

Together, these improvements yield significantly stronger performance for the legal domain. On BigLaw Bench, GPT-5 outperformed OpenAI’s previous frontier reasoning (o3) and non-reasoning (GPT-4.1) models by 5.09%, signifying a substantial improvement in legal reasoning capabilities.

Integrated into Harvey’s systems, these baseline capabilities can be leveraged to enable more powerful use cases in the document drafting and complex research domains. GPT-5 is also the first orchestration model that appears capable of combining these tasks—allowing for a single agent to both collaborate with a user on the research and produce the finished work product. For example, on a task like: Identify if any of these internal guidance documents are inconsistent with current regulation, we operate in the United States and the European Union . . . GPT-5 can be used to orchestrate agents that:

All while prompting the user as needed for additional context to ensure it reaches the goal as expected.

Coupled with our recently-announced data partnerships with LexisNexis and iManage, Harvey is now able to see the full picture – public and proprietary – before it acts. With GPT-5's substantially improved tool-use and drafting capabilities, we can now build a deeply integrated AI system that reasons over an organization’s internal data and leverages trusted third-party content in real-time.

Complex matters don’t unfold linearly; they advance dynamically through iteration, and in close collaboration with internal and external stakeholders. With GPT-5, and our product and data ingredients in place, Harvey’s north star of creating an intelligent coworker comes into focus.

An intelligent coworker does more than just respond to prompts; it understands context, anticipates needs, and manages complex tasks. It can plan a multi-stage project, execute steps using a variety of tools, and proactively surface insights or roadblocks for human review.

There are three pillars to this vision:

Collectively, these pillars represent a fundamental shift in how domain experts will interact with AI. While GPT-5 provides the raw potential, our core mission is to align these powerful new capabilities with the specific workflows, data sources, and standards of elite legal and professional services practice. This deep specialization is what converts raw model autonomy into reliable, practice‑ready expertise.

With the arrival of GPT-5, we’re moving beyond AI that simply executes commands to AI that orchestrates complex legal projects from start to finish. This new generation of AI doesn't just answer questions; it manages entire workflows, reasoning like a lawyer to navigate ambiguity, weigh competing arguments, and make strategic judgments.

This marks the transition from an era where lawyers use AI tools to one where they direct AI coworkers. It’s the foundation for a new kind of legal work, where your most complex processes are run by AI agents built on your expertise and guided by your strategic oversight.

To learn more about how Harvey empowers knowledge workers to solve their most complex problems, reach out for a demo.