AI TrackerGet the app
← All news
OpenAIModelsProducts

OpenAI releases an updated Astra for Law and opens access requests to all ChatGPT Enterprise customers

OpenAI releases an updated Astra for Law, a legal configuration of GPT-6.2 Astra that pairs the model with an expanded legal search index and instructions for researching a question, weighing authorities and writing the answer. First introduced in September to a small group of firms, it can now be requested by every ChatGPT Enterprise customer through the Trusted Access Program in the admin console; API access is by request through a form. In ChatGPT the model is gpt-6.2-astra-law. The results below are OpenAI's own, from third-party benchmarks it cites.

  • Harvey LAB-AA (Artificial Analysis's version of Harvey's Legal Agent Benchmark, 120 assignments across 24 areas of law): OpenAI says Astra for Law ranked first, earning full credit on 20.8% of assignments, ahead of GPT-6.2 Astra at 15.0% and Muse Wonder Full at 13.1%, with about 64,000 output tokens per assignment against 139,000 for Muse Wonder Full.
  • Hallucinations: it reports 0.03 material hallucinations per assignment, tied with GPT-6 for the lowest, against 0.59 for Claude Opus 5.5.
  • Vals AI Legal Research Bench: full credit on 112 of 208 questions (53.8%) in Vals AI's November 16 evaluation, up from about 46% in September.
  • LAB: USCL (Harvey and Grubel's brief-writing test): one of the top scores in both versions; with research it scored 1,844 against 1,817 for Opus 5.5, and the two were effectively even at 1,883 and 1,882 when given the precedents.

OpenAI says the legal search index has grown 34% since September to more than 310 million URLs, adding jurisdictions such as the UK, Australia and Canada. It quotes Harvey's head of legal research and a Sullivan & Cromwell partner as early users.

Get news like this as a push.

AI Tracker alerts you when something changes for the tools you use.

Get the app