Harvey turns legal context into stronger drafts with GPT-6 Astra
Harvey, a company that helps law firms and in-house legal teams securely deploy AI across complex legal workflows from litigation to mergers, says it is now using OpenAI's GPT-6 Astra model to bring more context into its drafting process. According to an OpenAI customer-story post, Harvey uses the model to help lawyers analyze, synthesize and draft from the material that shapes a legal matter, including court information, law firm documents, case law research and other sources of legal context. Harvey reports that, compared with other models, GPT-6 Astra has brought substantial improvements in document formatting and context awareness, helping customers get more complete documents that better reflect the underlying material, though no specific figures are given to quantify those gains. The post also describes Harvey's memory panel, a feature that brings individual lawyer preferences directly into the drafting workflow: a lawyer can encode preferences such as using numbered lists, prioritizing EDGAR as a source, or color-coding issues by priority. Those preferences appear alongside the source material and the draft memorandum, giving the lawyer a clearer way to guide the output. Because GPT-6 Astra can process more context, Harvey says it can produce higher-quality legal documents while its customers spend more of their time on strategy rather than assembly. The post includes an unattributed line describing the shift: "We can give more context to the model and produce better and better structured outputs."
Key facts
- Harvey uses GPT-6 Astra to draft from court information, law firm documents, case law research and other legal context sources.
- Harvey reports substantial (unquantified) improvements in document formatting and context awareness compared with other models.
- A memory panel lets lawyers encode preferences, such as numbered lists, prioritizing EDGAR as a source, or color-coding issues by priority, directly into the drafting workflow.
- Preferences are shown alongside source material and the draft memorandum for clearer guidance.
- Harvey says the added context capacity lets customers spend more time on strategy rather than document assembly.
Why it matters
The account illustrates how a frontier model's larger context handling translates into a concrete professional workflow: rather than a generic capability bump, Harvey frames GPT-6 Astra's context capacity as directly improving the structure and completeness of legal documents drafted from large volumes of case material.
Who it affects
Lawyers and in-house legal teams at law firms and companies that use Harvey for litigation, mergers and other complex legal workflows are the direct users; more broadly it signals how legal-tech vendors are positioning newer OpenAI models for document-heavy professional work.
How to use it
Through Harvey's memory panel, a lawyer can encode drafting preferences, such as using numbered lists, prioritizing EDGAR as a source, or color-coding issues by priority, and those preferences appear alongside the source material and the draft memo during the drafting process.
How solid is it
This is a customer case study published by OpenAI itself, so it is promotional in nature. It gives no numbers to back the claimed 'substantial improvements,' no date for when Harvey adopted the model, and does not name the individual quoted about giving more context to the model.
Risks and caveats
Beyond the lack of quantified metrics or an independent source, the account rests on a single vendor's description of its own product, with no named individuals at either Harvey or OpenAI to attribute the specific claims to.
“We can give more context to the model and produce better and better structured outputs.”
— Harvey, quoted in the OpenAI customer story