OpenAI Launches ChatGPT for Financial Services with Built-In Data and GPT-6 Astra
OpenAI has introduced ChatGPT for Financial Services, a dedicated workplace edition that couples GPT-6 Astra reasoning with built-in financial market data from PitchBook, LSEG News, and Daloopa. Designed alongside Morgan Stanley and Evercore, the platform automates equity research, valuation models, and client pitchbooks while eliminating manual data connectors.
What’s New
- Combines GPT-6 Astra reasoning with indexed market data from Daloopa, PitchBook, LSEG News, and Crunchbase.
- Co-developed with investment banks Morgan Stanley and Evercore to streamline equity research and valuation modeling.
- Achieves 69.9 percent on OfficeQA Pro Treasury bulletin analysis, compared to 60.2 percent for GPT-5.6 Sol.
- Integrates existing firm entitlements from S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, and Moody's via single sign-on.
- Generates client-ready pitchbooks, spreadsheets, and research notes directly in firm-approved Office templates.
- Protects sensitive data with encryption at rest and in transit, workspace retention controls, and audit log exports.
Why It Matters
For investment banks and financial analysts, natively indexed market datasets remove the daily friction of broken connectors and manual table re-entry. However, rolling out automated valuation models still requires strict administrative partitioning and verification gates to protect material non-public information.
OpenAI has launched ChatGPT for Financial Services, a dedicated enterprise tier tailored for investment bankers, equity research analysts, and institutional finance teams. The new release pairs the company's frontier reasoning model, GPT-6 Astra, directly with native data feeds from major financial information providers, tackling one of the most stubborn workflow bottlenecks in enterprise knowledge work.
Historically, generative AI models deployed in finance struggled with hallucinations on precise accounting disclosures or relied on brittle third-party Model Context Protocol connectors that required separate contracts and frequent maintenance. To solve this, OpenAI partnered with design collaborators Morgan Stanley and Evercore to overhaul the retrieval architecture. The platform indexes and hosts datasets directly from Daloopa, PitchBook, LSEG News, and Crunchbase on OpenAI infrastructure. Analysts evaluating corporate earnings can inspect granular citations that highlight the exact footnotes, balance sheet lines, or management commentary supporting figures such as adjusted EBITDA.
Under the hood, GPT-6 Astra delivers measurable gains in interpreting complex financial literature. On the OfficeQA Pro benchmark, which tests an agent's ability to navigate multi-column tables, charts, and technical annotations across United States Treasury Bulletins, GPT-6 Astra achieved a 69.9 percent accuracy score, surpassing the 60.2 percent recorded by GPT-5.6 Sol. Beyond document analysis, the model builds dynamic visualizations, including leveraged buyout value bridges, and formats outputs into custom corporate templates across Microsoft Excel, Word, and PowerPoint.
For institutions with existing subscriptions, OpenAI is collaborating with S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, and Moody's to recognize user entitlements through ChatGPT single sign-on. Over fifty additional connectors, including FactSet, Preqin, Datasite, Box, and Intapp, are supported out of the box.
Security controls mirror the standard set by ChatGPT Enterprise. Business data is not used to train models by default, communications and storage are fully encrypted, and compliance administrators can configure workspace retention policies or export audit logs through the OpenAI Compliance Platform. Multiple isolated workspaces can be provisioned to enforce information barriers between research desks and deal teams.
ChatGPT for Financial Services is available immediately to eligible financial institutions, with onboarding managed through OpenAI account representatives. Exact per-seat subscription pricing was not disclosed in the launch announcement.
