OpenAI has released Astra for Law, its GPT-6 Astra model paired with a dedicated legal search index and custom instructions. The index covers more than 230 million URLs of US case law, statutes, regulations, court rules and administrative decisions. It goes first to selected law firms through Trusted Access in ChatGPT and Codex, with API access described as coming soon under the name gpt-6-astra-law. Pricing is not disclosed. Named partners include Harvey, Legora, Sullivan & Cromwell, Latham & Watkins and Wachtell.
The headline figures come from OpenAI’s own testing on 200 questions drawn from the private validation set of Vals AI’s Legal Research Bench. Astra for Law passed 54.0% of them against 38.7% for GPT-6 Astra using web search alone, found 24% more reference cases on case-law questions, and retrieved up to 54% more relevant passages from the correct opinions.
Worth being clear about what that is. OpenAI ran the evaluation, on a set the public cannot see, and does not say who defined correctness or whether anyone replicated it. The single failure example on the page is also OpenAI’s account of a rival model returning a holding reversed on appeal. Legal research is a field where a confident wrong citation has already earned lawyers sanctions, so the distance between 54% and dependable is the number that matters.