OpenAI’s API price list now has a row for GPT-6 Astra for Law
The developer pricing page lists gpt-6-astra-law at $10 per million input tokens, $50 per million output tokens, and $1 per million cached input tokens, with promotional pricing at least through 31 December. Those three numbers match GPT-6 Astra’s short-context rates. A product post dated 17 September already described the configuration. The help center still says the API is coming soon.
Deep dive5 minIn-depth
Maya Chen5 min read
- OpenAI
- GPT-6 Astra
- Law
- API

Photo: OpenAISource page
OpenAI’s public API price list now includes a legal configuration, gpt-6-astra-law, at $10 per million input tokens, $50 per million output tokens, and $1 per million cached input tokens. The section is headed “Legal model configurations” and calls the row one of “our latest legal model configurations.” A line under the table says GPT-6 Astra for Law promotional pricing is available at least through 31 December 2026. The same page prices gpt-6-astra, for a prompt of up to 272,000 input tokens, at those same three rates. Astra’s long-context rates on that page are $20, $75, and $2. The legal table does not print a long-context price, or the cache-write prices listed for Astra.
Neural Brew, in a post by Lior timestamped 11:09 CET on Saturday, reported the row on the pricing page and said a model-card address it expected to find returned a 404, with no changelog entry alongside it. OpenAI’s own post, “Introducing Astra for Law,” is dated 17 September 2026. The help-center page for the product, checked Saturday, showed an update of three days earlier. It still says early access is for selected law firms in the United States, through Trusted Access and Codex, and that the API is coming soon. The model picker name on that page is GPT-6 Astra Law. Access, it says, is for eligible lawyers and people working under their supervision.
A search index, and a score against web search
The September post describes Astra for Law as GPT-6 Astra with settings, tools, and context for professional legal work. The tool it spends the most time on is a legal search index: U.S. case law, statutes, regulations, court rules, and administrative decisions, across a corpus of more than 230 million URLs, with sources added daily. OpenAI says work with the Free Law Project, the nonprofit behind CourtListener, brings that project’s case-law collection, covering more than 99.9 percent of published U.S. precedential case law, into the research setup. The help center repeats the categories and the daily updates, and tells readers to review answers and cited sources before relying on them.
The measurement in the September post is OpenAI’s. On 200 U.S. legal-research questions from the private validation set of Vals AI’s Legal Research Bench, at the highest reasoning effort for both systems, Astra for Law passed the evaluation’s overall correctness check on 54.0 percent of questions. GPT-6 Astra using web search alone passed on 38.7 percent. OpenAI calls that a 40 percent relative improvement, and says the legal configuration also produced more comprehensive answers. On case-law questions, it says Astra for Law found 24 percent more reference cases than GPT-6 Astra with web search, again at the highest effort. On an audited set of target passages, it says the legal setup retrieved up to 54 percent more relevant passages from the correct opinions, comparing the two systems at the same reasoning effort. The post does not print a score from an outside lab.
One worked example on the same page is OpenAI’s comparison, not a third-party test. The prompt is a misrepresentation claim about a retailer’s historical order average. The page shows Claude Fable 5.1 citing Paragon Metals Holdings v. Smith, a Delaware Superior Court case, and Astra for Law citing Oliver Wyman, Inc. v. Eielson, 282 F. Supp. 3d 684 (S.D.N.Y. 2017). OpenAI’s caption says that, given the same prompt, Astra for Law returned two closely matching precedents, and that in the litigation example Claude Fable 5.1 returned a holding that had been reversed on appeal, while in a transactional example it reported finding no such case. The page does not link the reversal.
In our early testing, Astra for Law showed strength across key aspects of legal research: grounding answers in on-point authorities, citing with precision, and offering practical, advisory guidance.
— Niko Grupen, head of applied research at Harvey, quoted in OpenAI’s 17 September post
Plugins, firm tools, and a price with no long-context column
The September post names Harvey and Legora as API customers that will be able to build on Astra for Law. It also describes work already done inside ChatGPT Enterprise with forward-deployed engineers. Sullivan & Cromwell built an agreement analyzer that applies the firm’s negotiating playbooks and selected precedents, then proposes redlines and draft client advice. Ropes & Gray built a deal-diligence system for a data room, including whether key customer contracts require notice or consent. Cooley built a tool, GO Public, for drafting an IPO filing and carrying a change in the deal across that filing. Skadden is quoted on tools for regulatory risk. Wachtell, Lipton, Rosen & Katz is quoted on conversations with the same engineering team. John Savva, a partner at Sullivan & Cromwell, is quoted on research depth in early previews. The quotes are the firms’ lines in OpenAI’s post.
The same post says OpenAI was launching 26 partner-built plugins that day. The examples it gives: iManage, so a lawyer can draft a negotiation brief in ChatGPT and save it to the matter file; Intapp, for activities that may need a time entry; DeepJudge, to bring prior deals into a comparison; Thomson Reuters, bringing HighQ matter context into ChatGPT and previewing a CoCounsel Legal connector. Joel Hron, Thomson Reuters’ chief technology officer, is quoted saying CoCounsel remains the firm’s own system and that the OpenAI work is meant to make those capabilities available in the environments customers choose. Nine community plugins, from lawyers and legal engineers at LegalQuants, LECG, and Skills.law, come with 47 custom skills. The post also says ChatGPT for Word was becoming generally available that day, for proofreading, suggested edits, and formatting flags. Those dates belong to 17 September. Saturday’s pricing page does not repeat them.
On controls, the September post says eligible firms get a Trusted Access program, Zero Data Retention on the API, and ChatGPT Enterprise use excluded from human review by default. It says OpenAI is working with Latham & Watkins on information permissions, ethical walls, client instructions, and firm oversight. Michael Rubin, chair of Latham’s AI strategy committee, is quoted on governance and accountability. The enterprise rate card, a separate help-center page, lists “GPT-6 Astra Law” in the ChatGPT Work and Codex table at $10, $1, and $50 per million tokens, the same three numbers as the API row. A footnote there says access requires approval through Trusted Access, and that promotional pricing runs at least through 31 December 2026. The API price table itself does not say who can call gpt-6-astra-law, or whether the endpoint is open. The help center still says the API is coming soon.


