Hebbia sees improvements in financial research with GPT-6 Astra

What happened
Hebbia has added OpenAI's GPT-6 Astra model to its Matrix platform to help bankers conduct financial research, especially when evaluating acquisitions. The integration aims to improve the accuracy of source citations and automate the creation of presentations based on bankers' instructions. Hebbia tested GPT-6 Astra in August and September 2026 using real financial research scenarios and found it performed better than other models in both citation relevance and presentation quality. The platform now allows users to trace answers back to original documents and generate slides that match firm branding and requirements.
Key details
- GPT-6 Astra scored 13.6 percentage points higher than the next best model on citing required source types.
- 99.5% of excerpts cited by GPT-6 Astra were relevant to the research questions.
- In presentation creation, GPT-6 Astra met 97.9% of requirements, 14.4 percentage points ahead of alternatives.
- Testing was conducted with a prerelease build in August-September 2026 using domain expert-reviewed criteria.
- Matrix users can view citations directly in documents and follow slide footnotes back to sources.
Why it matters
Bankers and financial analysts using Hebbia's Matrix platform can now spend less time checking sources and reformatting presentations, as the AI provides more accurate citations and higher-quality slides. This reduces manual work and the risk of errors in high-stakes financial decisions. Firms benefit from greater traceability and confidence in AI-generated insights, which can speed up deal assessments and improve the reliability of research outputs.
