OpenAI has published an API price for the Astra for Law configuration.
Developers need to separate a priced model record from confirmed general availability and professional suitability.
What happened?
OpenAI's current API pricing page lists gpt-6-astra-law under legal model configurations at $10 per million input tokens, $50 per million output tokens and $1 per million cached input tokens. The same row says promotional pricing is available at least through December 31, 2026. This is an official pricing record, not a leaked screenshot or a reseller claim.
The configuration was announced earlier as GPT-6 Astra tailored for professional legal work, with legal search, instructions, tools and privacy/governance controls. OpenAI's launch announcement said selected law firms would initially receive access through Trusted Access in ChatGPT and Codex and that API access was coming soon. The current pricing row shows that OpenAI has published prices; it does not by itself show that every API customer can call the model today.
OpenAI's own help text frames the system as assistance for professional legal work and tells users to review answers and cited sources. That warning remains relevant even if the API row becomes generally callable: a model configuration can be priced and technically reachable without being a substitute for legal research, advice or attorney review.
Price is not the same as availability
The pages reviewed do not establish a public endpoint name beyond the pricing configuration, a general-access date, regional eligibility, rate limits, quota, approval process or whether the legal configuration is visible in every account. The help page also tells users to review answers and cited sources before relying on them, which is a product caveat rather than a guarantee of legal accuracy.
The promotional wording is “at least through” December 31, not a promise that the price ends on that date. Nor does the token price capture retrieval, tool, storage, review or professional-liability costs. A legal workflow may pay for model output while still needing a licensed professional to validate the result.
The legal configuration's label also should not be confused with a new model family. OpenAI describes Astra for Law as GPT-6 Astra adapted with legal-oriented settings and tools, while the pricing table places it under configurations. The listed cached-input rate may matter for repeated document context, but actual bills depend on cache eligibility, request design, tool calls and the amount of material sent. The launch and help pages emphasize review of answers and cited sources. No source reviewed supplies an independent legal benchmark, malpractice protection, jurisdiction coverage or a promise that citations are complete.
Source: OpenAI ↗ · Source: OpenAI ↗ · Source: OpenAI ↗ · Source: OpenAI ↗
Why does it matter?
Publishing the price before broad access is a useful signal for developers planning a legal workflow, but it also creates a trap: a price table can look like a launch announcement even when the access path remains selected or gated. Teams should confirm that their organization can actually reach the model, identify which endpoint and limits apply, and budget for cited-source review before migrating a legal task.
The practical distinction is between “priced configuration” and “available service.” OpenAI may be preparing the API surface while preserving a controlled rollout, and the docs may change before the stated promotion ends. Readers should treat the official row as the source of truth for the displayed rates at the time checked, not as evidence that the model has passed an independent legal benchmark or is suitable for unsupervised advice.
For legal-technology teams, the first question is operational rather than promotional: can the organization obtain access and constrain the workflow to tasks a reviewer can validate? A lower token price does not remove confidentiality, privilege, retention or professional-responsibility decisions. Teams should test retrieval quality on their own approved corpus, log citations, set escalation rules and compare total workflow cost instead of comparing only the output rate. The article reports a price record, not a recommendation to delegate legal judgment.
Source: OpenAI ↗ · Source: OpenAI ↗ · Source: OpenAI ↗ · Source: OpenAI ↗ · Source: OpenAI ↗
Read the sources
- OpenAI API pricing ↗
OpenAI - Astra for Law announcement ↗
OpenAI · September 17, 2026 - Astra for Law help ↗
OpenAI - OpenAI API models overview ↗
OpenAI - OpenAI API rate limits ↗
OpenAI
