What changed

On November 17, 2026, OpenAI announced the release of an updated Astra for Law model, which is a configuration of their latest GPT-6.2 Astra model. This version has achieved leading marks on multiple legal benchmarks, including the Harvey LAB-AA and Vals AI’s Legal Research Bench. All ChatGPT Enterprise customers can now request access to this model through the Trusted Access Program, including via the API.

What this means in practice

Astra for Law is designed to assist legal professionals by combining GPT-6.2 Astra with an expanded legal search index and specific instructions for researching, weighing authorities, and writing legal answers. The model earned full credit on the Harvey LAB-AA tests nearly 40% more often than the runner-up, GPT-6.2 Astra. On Vals AI’s Legal Research Bench, the share of research answers that passed every check rose from about 46% to 53.8% in less than two months.

The model also demonstrated a significant reduction in hallucinations, with Astra for Law tying with GPT-6 for the lowest rate of material hallucinations at 0.03 per assignment, compared to 0.59 for Claude Opus 5.5. Additionally, Astra for Law achieved a top score with 54% less output per assignment than the highest-scoring non-OpenAI model, averaging about 64,000 output tokens per assignment.

Limitations and availability

The updated Astra for Law model is available to all ChatGPT Enterprise customers through the Trusted Access Program. The model is based on GPT-6.2 Astra and includes an expanded legal search index covering over 310 million URLs, including international jurisdictions such as the UK, Australia, and Canada. The model’s performance on legal benchmarks is reported by OpenAI, but independent testing has not been conducted. The model is not yet generally available to the public and requires access through the Trusted Access Program.

Source facts for OpenAI releases Astra for Law 6.2 with improved legal benchmarks: November 17, 2026CompanyAstra for Law leads in legal intelligenceOur latest model configuration earned leading marks on multiple benchmarks. LAB-AA⁠(opens in a new window) is Artificial Analysis’s version of that test. For legal teams paying by usage, this can mean the difference between staying within budget and blowing through it.Each dot is a model configuration.
Original AINewsTrend explainer based on the cited source. This graphic summarizes source statements; it is not an independent benchmark.

Sources and further reading

  1. openai: Astra for Law leads in legal intelligence