Hebbia sees improvements in financial research with GPT-6 Astra

OpenAI
Hebbia tested GPT-6 Astra for financial research, finding it excelled at source citation and presentation quality with high accuracy.

Summary

Hebbia, a financial research platform, tested OpenAI's GPT-6 Astra to improve two key tasks: finding sources for questions and building presentations that follow banker briefs. In tests using a hypothetical data room with roughly 2,000 files, GPT-6 Astra selected relevant documents and Matrix cells to support answers, then generated native PowerPoint slides following user structure and guidelines. The model scored 13.6 percentage points above the best alternative on citing required source types, with 99.5% of cited excerpts relevant to the question. It also met 97.9% of applicable presentation requirements, 14.4 percentage points ahead of the best alternative, and ranked first on both following the brief and presentation quality. Hebbia emphasizes traceability through citations and slide footnotes that link back to source material. Kavita Anand, Member of Technical Staff, Product at Hebbia, noted that GPT-6 Astra is the first model to rank first on both instruction following and craft, potentially reducing hours spent reformatting decks and checking sources. She also envisions a future role for agents that monitor information in the background and surface signals for finance professionals.

(Source:OpenAI)