ChatGPT for Financial Services

ChatGPT for Financial Services is OpenAI's vertical workspace for investment banking and equity research, announced September 10, 2026. It is a tailored ChatGPT Work experience rather than a new model: GPT-6 Astra does the reasoning, and the differentiation sits in the data, the connectors, and the governance wrapped around it. OpenAI built it with Morgan Stanley and Evercore as design partners and says the starting scope is deliberately narrow, covering the two workflows those teams named as the biggest pain points.

Key features

• Premium data included, not negotiated: datasets from Daloopa, PitchBook, LSEG News, Crunchbase, and Fiscal.ai cover earnings transcripts, financial statements, company fundamentals, and private companies. OpenAI indexes and hosts this data itself, which is what enables granular citations that trace a figure back to a specific table or passage.
• Entitlement-aware access: firms that already pay for S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, or Moody's can have users recognized through their ChatGPT sign-in, so existing entitlements carry over instead of being repurchased.
• Optimized MCP connectors: the most common financial data MCPs are tuned for immediate use, and the broader connector ecosystem runs past 50 integrations including Datasite, Box, Preqin, FactSet, and Intapp.
• Firm templates for artifacts: administrators publish Excel, Word, and PowerPoint templates from a dedicated admin page, so output arrives as valuation models, research notes, and pitchbooks in the firm's own format and style.
• Information barriers: multiple workspaces, role-based access controls, SAML SSO, and SCIM provisioning, with business data excluded from training by default and logs exportable through the OpenAI Compliance Platform.

Who should use it

• Investment banking and equity research teams doing value analysis, LBO modelling, buyer screening, earnings analysis, and pitchbook preparation.
• Firms already on ChatGPT Enterprise that want the finance-specific data layer and admin controls without a separate procurement cycle.
• Compliance and IT leads who need role-scoped access and an audit trail before they will let analysts near an AI workspace.

What OpenAI claims, and what to check

OpenAI reports GPT-6 Astra at 69.9% on OfficeQA Pro, which tests whether an agent can find and analyse information across US Treasury Bulletins with complex tables and footnotes, against 60.2% for GPT-5.6 Sol. That is a first-party benchmark on a task chosen to look like the product's day job, so treat it as a vendor number and run your own documents through it.

Pricing and availability

There is no public price list. ChatGPT for Financial Services is available to eligible financial institutions through sales, which in practice means a negotiated enterprise agreement rather than self-serve signup.

Alternatives

Anthropic ships a comparable vertical with Claude for Financial Services, an open plugin pack of reference agents and skills for DCF/LBO models and earnings review. The trade-off is real: OpenAI bundles licensed data and hosted retrieval, while Anthropic's pack is portable and works against whatever data you already have.

The practical read is that the model is not the moat here. The moat is the licensed data, the citations, and the admin controls, and those are exactly the parts you cannot evaluate from a press release.