Alongside the public release of GPT-5, we are very excited to announce it is now available in Rogo. As OpenAI's premier partner for finance, we've spent significant time over the past month testing GPT-5 against core financial workflows.

Integrating GPT-5 into our response engine has driven a step change in Rogo’s Answer Quality score — a proprietary benchmark we fine-tune across hundreds of investment banking workflows, from M&A comps to full slide deck generation.

This launch follows extensive tuning and evaluation on Rogo’s financial reasoning and QA dataset, which spans over 100 core use cases bankers and investors rely on daily. These include precision-heavy tasks like valuation analysis, precedent transactions, and pitch deck generation, all designed to push LLMs on accuracy, nuance, and reasoning quality.

Rogo users will see immediate gains in answer precision, contextual nuance, and multi-step reasoning. Our collaboration with OpenAI on this launch sets a new standard for AI-powered finance tools. We tested GPT-5 on two criteria — Win Rate and Source Attribution — and see meaningful improvements.

Win Rate

We found that GPT-5 significantly outperforms o4-mini in head-to-head evaluations conducted by custom-trained models built to assess helpfulness, source credibility, formatting quality, and conciseness. GPT-5 wins 62% of matchups on low- to medium-complexity questions, and 70% on highly complex reasoning tasks.

Source Attribution (AIS)

Source Attribution evaluates how well-grounded an AI-generated response is. It uses a custom trained model to identify claims needing citations and assesses the accuracy and completeness of the sources provided. This metric also captures the hallucination rate. We see a meaningful improvement in our Source Attribution Score with GPT-5.

To see GPT-5 live in Rogo today, reach out to us at sales@rogo.ai.