OpenAI Launches GPT-6 Astra for Law to Target Legal Research Market

OpenAI has officially launched a legal industry variant of its flagship model titled Astra for Law, providing law firms and legal software providers with dedicated capabilities for complex research, document drafting, and custom workflow development. The new platform pairs the GPT-6 Astra architecture with a specialized Legal Search Index containing more than 230 million URLs covering US case law, statutes, regulations, court rules, and administrative decisions.

The commercial rollout places OpenAI in direct competition with rival artificial intelligence labs seeking to secure enterprise contracts across the professional legal sector. Competitors including Anthropic and Google have recently expanded their own legal-focused enterprise offerings and agent integrations, while legal publisher Thomson Reuters rolled out its open-source-backed Thomson 1.0 model last month.

To evaluate the system, OpenAI tested Astra for Law on 200 questions derived from the private validation set of Vals AI's Legal Research Bench. According to the company, the legal variant achieved a 54.0 percent correctness score at its highest reasoning effort, compared to 38.7 percent for standard GPT-6 Astra utilizing web search alone, representing a reported 40 percent relative improvement.

The underlying search infrastructure incorporates case-law collections from the nonprofit Free Law Project, the organization behind CourtListener, which covers more than 99.9 percent of published US precedential case law. OpenAI reported that the legal configuration retrieved 24 percent more reference cases on case-law queries and up to 54 percent more relevant target passages from correct court opinions than the unconfigured flagship model.

Recognizing strict security requirements across the legal industry, OpenAI structured the platform with specialized data protections. Eligible law firms participating in its Trusted Access program receive Zero Data Retention on API endpoints, and ChatGPT Enterprise usage is excluded from human review by default.

The company collaborated with law firm Latham & Watkins on governance architecture, addressing information permissions, ethical walls, client instructions, and internal oversight. Additional early adopters and testing partners include Sullivan & Cromwell, Ropes & Gray, Cooley, and Wachtell Lipton, which developed specialized internal use cases ranging from agreement analyzers to deal diligence systems and IPO preparation tools.

The platform debuted with native support from 26 partner-built plugins and 47 community-contributed tools designed to bridge ChatGPT with existing legal software environments. Integrations include platforms such as Relativity, Clio, iManage, Intapp, and Thomson Reuters, allowing firms to connect the model directly into established matter management and document review software.

Legal AI vendors Harvey and Legora were named as primary API customers building commercial applications on top of the infrastructure. Harvey, which partnered closely with OpenAI on custom training methodologies, reported that legal professionals in early evaluations favored specialized legal model outputs for thoroughness and source grounding, while noting that ongoing rollout details including official API pricing and general availability schedules remain unannounced.