Hebbia sees improvements in financial research with GPT-6 Astra

OpenAI
Hebbia tested OpenAI's GPT-6 Astra on financial research tasks, finding it outperforms alternatives in source citation and presentation generation.

Summary

Hebbia, a financial research platform, tested OpenAI's prerelease GPT-6 Astra model on two core tasks: locating supporting sources and building client-ready presentations. In tests conducted in August-September 2026 using the same financial research cases across models, GPT-6 Astra scored 13.6 percentage points above the best alternative on citing the required types of sources, 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, ranking first on both brief adherence and presentation quality. Kavita Anand, a member of Hebbia's product technical staff, noted that previous models usually excelled on one dimension or the other, while GPT-6 Astra leads on both. Hebbia's Matrix platform links answers to underlying documents for traceability, and GPT-6 Astra generates native PowerPoint slides following user templates and guidelines. Anand said the model could reduce the time bankers spend reformatting decks and checking sources, turning overnight work into minutes of review. She also sees a future role for agents that monitor information in the background and surface signals for finance professionals.

(Source:OpenAI)