OpenAI's GPT-6 Astra Targets Wall Street With Built-In Premium Market Data

OpenAI has rolled out a version of ChatGPT tuned for investment bankers and equity research analysts, bundling premium market data, the new GPT-6 Astra reasoning model, and firm-branded output templates into a single workspace. ChatGPT for Financial Services is designed to handle the workflows that dominate a banker's day: valuation models, buyer screens, earnings analysis, and pitchbook prep.

The product was shaped by a design partnership with Morgan Stanley and Evercore, whose analysts helped OpenAI identify where frontier models get stuck in real financial work. The answer, unsurprisingly, was data access and generating polished deliverables that match the firm's house style.

Data pipes, pre-plumbed

The biggest structural change is that OpenAI is now hosting premium financial datasets directly on its infrastructure rather than expecting every firm to wire up its own connectors. The product ships with feeds from Daloopa, PitchBook, and LSEG News covering earnings transcripts, financial statements, company fundamentals, and private companies, with no separate contracts to negotiate.

For firms that already pay for other terminals, OpenAI is working with S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, and Moody's on shared sign-in and entitlement integrations so users get automatic access to what they're already licensed for. An MCP connector ecosystem adds more than 50 integrations including Datasite, Box, Preqin, and Intapp, with the most-used ones like S&P Global and FactSet specifically tuned for reliability through automated evaluation loops.

What GPT-6 Astra brings to the desk

GPT-6 Astra is positioned as state of the art across three capabilities that matter for this domain: information retrieval, financial reasoning, and artifact generation. OpenAI points to a new benchmark called OfficeQA Pro, which tests whether AI agents can find and analyze information across U.S. Treasury Bulletins, including complex financial tables, charts, and supporting footnotes. GPT-6 Astra scores 69.9%, compared with 60.2% for GPT-5.6 Sol.

In practice, the model is meant to navigate figures, tables, and supporting notes in financial documents, run financial analysis and draw conclusions, then synthesize the result into documents, spreadsheets, and slides. That last step is where the template system kicks in.

Firm templates and traceable citations

Administrators can publish Excel, Word, and PowerPoint templates through a dedicated admin page, so the models, research notes, and pitchbooks generated by the tool come out in the firm's own formatting. Every output ties back to sources: analysts can trace figures and claims to specific paragraphs and tables, and preview the supporting passage from a citation before committing anything to a client deck.

The example OpenAI walks through is telling. A banker running a P&L normalization analysis can inspect the reconciliation and notes behind an adjusted EBITDA figure, see which costs were excluded, and decide how to use it in a valuation. That kind of source-anchored reasoning is the difference between a tool a compliance team will greenlight and one that stays banned on the trading floor.

Security posture and governance

Given the sensitivity of MNPI and client confidentiality, the product inherits ChatGPT Enterprise's guarantees:

• SAML SSO, SCIM provisioning, and role-based access controls
• Business data not used to train models by default
• Encryption at rest and in transit, plus configurable workspace retention
• Compliance teams can export supported workspace logs through the OpenAI Compliance Platform into their existing audit and investigation workflows
• Multiple workspaces to enforce information barriers between deal teams

Following Palantir into the pit

OpenAI is clearly running the vertical SaaS playbook that Palantir, Hebbia, and Rogo have been executing in finance, with two structural advantages: the underlying model and direct data-provider deals. The wedge is investment banking and equity research, where reliable data access and polished artifact creation were the pain points partners flagged as biggest.

The competitive pressure lands on a few groups. Bloomberg's GPT effort and startup research copilots now face a distribution giant with native premium data. Data vendors that sign on, Daloopa and PitchBook among them, get placement inside the workflow layer where analysts actually spend their time, which is worth more than API revenue. Vendors that don't sign risk being routed around.

For anyone building in this space, the practical takeaway is that the moat is shifting from model quality to data partnerships and workflow integration. If your product depends on scraping filings or reformatting Excel, that surface is now table stakes inside an enterprise SKU. The interesting build targets move up the stack: proprietary datasets, deal-team collaboration primitives, and the compliance tooling that determines whether a bank will actually deploy any of this. ChatGPT for Financial Services is available now to eligible financial institutions through OpenAI's sales team.