What is Astra for Law?

Astra for Law is OpenAI's legal configuration of GPT-6 Astra, announced September 17, 2026. It combines the model with a new legal search index covering more than 230 million URLs of US case law, statutes, regulations, court rules and administrative decisions, plus built-in instructions for legal analysis and writing and firm-level controls. It appears in the ChatGPT model picker as 'GPT-6 Astra Law' and will reach the API as gpt-6-astra-law. Selected law firms get it first through a Trusted Access programme in ChatGPT and Codex.

How accurate is Astra for Law?

On 200 questions from the private validation set of Vals AI's Legal Research Bench, Astra for Law passed the overall correctness check on 54.0% of questions at the highest reasoning effort, versus 38.7% for GPT-6 Astra using web search alone — a 40% relative improvement. It found 24% more reference cases and retrieved up to 54% more relevant passages from the correct opinions. The testing is OpenAI's own and has not been independently audited; the system still fails the correctness check on nearly half the questions.

Where does the Astra for Law index get its case law?

Most of the case law comes from the Free Law Project, the nonprofit behind CourtListener, whose collection covers more than 99.9% of published US precedential case law. OpenAI adds sources daily and positions the index as complementary to licensed content from Thomson Reuters (Westlaw) and similar providers rather than a replacement.

How much does Astra for Law cost?

OpenAI has not published Astra for Law pricing as of September 19, 2026. It is offered through Trusted Access for eligible law firms inside ChatGPT Enterprise and Codex, with Zero Data Retention on the API and human review excluded by default. The underlying GPT-6 Astra API model costs $10 per million input tokens and $50 per million output tokens; expect the law variant to be priced at or above that when the API opens.

Quick Answer

What Is OpenAI Astra for Law? GPT-6 Legal Index Explained

Published: September 19, 2026

What Astra for Law is

On September 17, 2026 OpenAI introduced Astra for Law: its flagship GPT-6 Astra model configured for professional legal work. Three things are bundled together:

A legal search index — more than 230 million URLs of US case law, statutes, regulations, court rules and administrative decisions, with sources added daily.
Instructions for legal analysis and writing — distinguishing a holding from dicta, addressing adverse authority, explaining how a contract exception shifts risk.
Controls for firms — a Trusted Access Program, Zero Data Retention on the API, ChatGPT Enterprise usage excluded from human review by default, and governance design work with Latham & Watkins on information permissions and ethical walls.

It shows up in the ChatGPT model picker as “GPT-6 Astra Law” and will reach the API as gpt-6-astra-law. As OpenAI’s frontier models advance, it says the legal capabilities will move to the newer models.

The index is the real news

The model configuration matters less than what sits under it. OpenAI now operates its own legal research corpus, and most of the case law comes from the Free Law Project, the nonprofit behind CourtListener, whose collection covers more than 99.9% of published US precedential case law.

OpenAI is careful to call the index complementary to “the licensed content and specialist products firms rely on from providers such as Thomson Reuters.” Thomson Reuters supplied a supportive quote for the launch, is bringing HighQ matter context into ChatGPT, and is previewing a CoCounsel Legal connector. It also owns Westlaw — the product an in-house OpenAI index most obviously competes with.

What the benchmark shows, and what it does not

OpenAI tested the full configuration on 200 US legal research questions from the private validation set of Vals AI’s Legal Research Bench:

Metric (highest reasoning effort)	GPT-6 Astra + web search	Astra for Law
Overall correctness check passed	38.7%	54.0% (+40% relative)
Reference cases found (case-law questions)	baseline	+24%
Relevant passages retrieved from correct opinions	baseline	up to +54%

Two caveats. The testing is OpenAI’s own, as is a published side-by-side against Claude Fable 5.1 in which OpenAI says the rival cited an overturned holding; no independent party has audited either. And 54% correct means the system fails nearly half the questions — a “supported answer” still needs a lawyer to read the authorities.

Who gets it and what it costs

Access: selected law firms first, through Trusted Access in ChatGPT and Codex; API “coming soon.”
Builders: Harvey (valued at $15.6B after a $550M round in September 2026) and Legora will build on Astra for Law through the API.
Price: unpublished. GPT-6 Astra itself is $10/$50 per million input/output tokens on the API (verified September 4, 2026), with cached input at $1.00 — a floor for what the law variant will cost.

The plugin ecosystem

Alongside the model OpenAI launched 26 partner-built plugins connecting ChatGPT to tools firms already use — Relativity, Clio, iManage, Intapp, DeepJudge, Litera, Everlaw, Consilio, Laurel and more — plus 9 community plugins from LegalQuants, LECG and Skills.law with 47 custom skills. ChatGPT for Word went generally available the same day. Read the partner press releases together and each describes where its own subscription stays: time entry continues in Laurel, deeper workflows in Litera, substantive analysis inside Relativity.

Firms already building custom tools with OpenAI’s forward-deployed engineers include Sullivan & Cromwell (agreement analyzer), Ropes & Gray (deal diligence), Cooley (GO Public for IPO filings) and Wachtell, Lipton, Rosen & Katz.

How it fits the market

Astra for Law lands four months after Anthropic’s Claude for Legal (May 12, 2026) and in the same month Harvey raised at $15.6B. Every frontier lab now sells directly to law firms while also selling infrastructure to the legal-AI startups that compete for the same firms. For a side-by-side of pricing, access and strengths, see Astra for Law vs Harvey vs CoCounsel vs Claude for Legal.