OpenAI launches ChatGPT for Financial Services with financial data sources

• OpenAI has shipped ChatGPT for Financial Services, a finance-specific ChatGPT Work experience pairing GPT-6 Astra with research, modeling, and client-material workflows, as OpenAI's announcement describes.
• Built-in data starts with Daloopa, PitchBook, and LSEG News, which OpenAI listed in its data update, and the interface can preview the passage behind a cited figure or claim, per OpenAI's evidence-preview post.
• Firms can generate editable models, research notes, and pitchbooks in their own Office templates, according to OpenAI's template post.
• A comparison chart shared by testingcatalog puts GPT-6 Astra at 69.9% correctness on OfficeQA Pro, versus 60.2% for GPT-5.6 Sol and 62.4% for Claude Fable 5.1.

OpenAI's launch post says it hosts and indexes the vendor data itself. Its new Financial Services Terms preserve data-provider restrictions even when Partner Data appears in output, while Reuters' report identifies Morgan Stanley and Evercore as design partners.

The tailored ChatGPT Work experience combines GPT-6 Astra with premium market and company data. OpenAI says it was shaped with Morgan Stanley and Evercore, and Reuters reports the initial audience is investment bankers and equity researchers.

OpenAI lists these built-in datasets:

• Daloopa
• PitchBook
• LSEG News
• Crunchbase, in the official launch post

The company says the data is indexed and hosted on its infrastructure. It is also working on shared sign-in and entitlement integrations for existing subscriptions with S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, and Moody's, according to the launch post.

Access is limited to eligible financial institutions, as testingcatalog's launch summary notes.

OpenAI's analysis interface traces figures and claims to specific paragraphs and tables, then lets a user preview the supporting passage without leaving the workstream.

That turns citations into a retrieval surface rather than a footnote. OpenAI attributes the feature to its hosted, indexed data layer in the official announcement.

The product generates editable financial models, research notes, and pitchbooks using a firm's own Excel, Word, and PowerPoint templates.

Administrators publish those templates and style guidance from a dedicated admin page, OpenAI says. The same announcement positions the artifacts alongside research across sources, financial-document reconciliation, interactive analysis, charts, and visualizations.

A chart circulated by testingcatalog gives the only supplied numerical model comparison:

• GPT-6 Astra: 69.9%
• GPT-5.6 Sol: 60.2%, a 9.7-point gap to Astra
• Claude Fable 5.1: 62.4%, a 7.5-point gap to Astra

The chart labels its vertical axis “Correctness” but does not describe OfficeQA Pro's dataset, task mix, or evaluation protocol.

OpenAI's Financial Services Terms say third-party Partner Data remains governed by each provider's terms. Those terms can limit copying, downloading, storage, model training, derived-data creation, or redistribution, including when the data is incorporated into output.

Exporting or sharing does not expand those rights, and a customer's ownership of output does not include Partner Data. The consequential detail is contractual: the product's generated documents can carry data-license boundaries with them.

The terms also characterize the service and its output as informational tools, not financial or investment advice.