Astra for Law: A GPT-6 Model Built for Legal
OpenAI has entered the legal-tech arena in earnest, unveiling Astra for Law, a configuration of its GPT-6 Astra model purpose-built for legal research, analysis and drafting. The launch pairs the model with a dedicated legal search index and 26 partner plugins from vendors including Thomson Reuters, Intapp, Relativity, Clio and DeepJudge, alongside nine additional community-built plugins carrying 47 custom skills. Harvey and Legora are named as early API customers, building Astra for Law into their own products and workflows.
What's New
Astra for Law is not a standalone model but GPT-6 Astra wrapped in legal-specific infrastructure: a search index spanning US case law, statutes, regulations, court rules and administrative decisions across more than 230 million URLs, built on data from the non-profit Free Law Project's CourtListener database.
On Vals AI's Legal Research Bench, OpenAI reports meaningful gains over GPT-6 Astra using web search alone, roughly a 40 per cent improvement in correctness and about 24 per cent more relevant cases surfaced on case-law-focused questions. No independent benchmark yet compares performance against incumbent legal research platforms.
Access is currently limited to a Trusted Access Program within ChatGPT and Codex, with an API version expected later. OpenAI is offering eligible firms zero data retention and default exclusion from human review, with Latham & Watkins advising on governance, including information permissions, ethical walls, client-instruction handling and firm oversight.
Firms Are Already Building on It
Several early-adopter firms have used embedded OpenAI engineering support to build custom tools on ChatGPT Enterprise:
- Sullivan & Cromwell built an agreement analyser that draws on the firm's negotiating playbooks to generate proposed redlines.
- Ropes & Gray built a diligence system structured around how its lawyers work through a data room.
- Cooley built 'GO Public', a tool supporting IPO preparation, including drafting assistance for the filing itself.
Sullivan & Cromwell partner John Savva described the models as showing 'impressive research depth and sensitivity to authority.'
Vendors Frame It as Additive, Not a Replacement
Thomson Reuters, whose HighQ matter context and forthcoming CoCounsel Legal connector are among the launch's 26 plugins, carefully positioned connectivity as only part of the value proposition. Chief technology officer Joel Hron said legal professionals need more than access to information; they need 'trusted intelligence, relevant enterprise and matter context, purpose-built legal capabilities, and the governance required for high-stakes work.'
Why It Matters
Astra for Law follows Anthropic's Claude for Legal, which launched in March, intensifying competition among frontier AI labs for the legal vertical. It also arrives roughly two weeks after GPT-6 Astra's broader rollout and about a week after OpenAI's ChatGPT for Financial Services launch, part of a wider push into industry-specific products.
Share
See more LPI in Google Search and AI Overviews: add LPI as a preferred source
Disclaimer: The information in this article is general in nature and does not constitute legal, financial or other professional advice. It is not a substitute for advice tailored to your circumstances. While Legal Practice Intelligence takes care to ensure content is accurate at the time of publication, we make no warranty as to its completeness or currency and accept no liability for any loss arising from reliance on it. Views expressed by contributors are their own and do not necessarily reflect those of Legal Practice Intelligence. References to products, services or vendors do not constitute an endorsement, and some content may be produced in partnership with, or sponsored by, organisations mentioned.
