OpenAI GABRIEL — Turning Qualitative Research Into Data at Scale

The Problem With Qualitative Data

Qualitative data tells the richest stories about the world — what people say, write, teach, argue, and experience. It spans everything from syllabi and interviews to social media posts and photographs. There is a tremendous amount of it. But transforming that type of data into rigorous, analyzable evidence is incredibly time-consuming. A researcher studying political discourse might spend months manually coding thousands of speeches. A sociologist analyzing survey responses faces the same bottleneck at every scale.

This is the gap OpenAI is now addressing with GABRIEL.

What GABRIEL Is

OpenAI's Economic Research Team has released GABRIEL — an open-source toolkit that uses GPT to turn unstructured text and images into quantitative measurements. It is designed for economists, social scientists, and data scientists to study qualitative data at scale. Released on February 13, 2026, it is free to use and openly available to the research community.

The toolkit bridges one of the most persistent gaps in social science methodology: the ability to extract structured, comparable, statistically analyzable signals from the messy, context-rich data that defines human experience.

What It Can Do

GABRIEL allows researchers to:

• Convert text into numbers — turning open-ended survey responses, interviews, news articles, or social media into coded, quantifiable variables
• Analyze images at scale — extracting measurable features from photographs, diagrams, or visual content that would otherwise require manual review
• Study qualitative data rigorously — applying GPT's language understanding to produce measurements comparable across thousands or millions of data points
• Work across domains — applicable to political science, sociology, economics, education research, public health communication, and more

Traditionally, this kind of coding work was done by hand — by teams of research assistants applying predefined frameworks to each document, one at a time. GABRIEL compresses that process from months to hours.

Why This Matters for Social Science

Social science has always been data-rich but analysis-poor when it comes to qualitative material. The gap isn't the lack of interesting questions — it's the cost of answering them at the scale needed for statistical power. A study examining how university syllabi have changed over a decade across hundreds of institutions used to require a small army of coders. GABRIEL makes that kind of project tractable for a single researcher or small team.

GPT-5.2 scored 92% on the GPQA benchmark — a doctoral-level scientific knowledge test where human experts average around 70%. That's not a language parlor trick. At that level of comprehension, the model can apply nuanced, context-aware coding frameworks to complex qualitative material in ways that were simply not possible with earlier NLP tools.

Open Source by Design

Releasing GABRIEL as an open-source toolkit is a deliberate choice. OpenAI's broader mission includes enabling scientists to move faster and solve harder problems — and locking a research tool behind a paywall would contradict that directly. By making GABRIEL freely available, the Economic Research Team is betting that broad adoption by the research community will generate more value — through methodological development, peer review, and real-world applications — than any proprietary model could.

It also positions OpenAI as a genuine partner to academia rather than simply a vendor, which matters as universities and research institutions navigate their own AI adoption decisions.

Part of a Larger Science Push

GABRIEL doesn't exist in isolation. It's part of OpenAI's broader strategy to become the foundational AI layer for scientific research — alongside Prism for scientific writing, ChatGPT Health for medicine, and OpenAI for Healthcare for clinical institutions. The pattern is consistent: identify a domain where AI can reduce friction in daily research work, build a purpose-fit tool, and release it with low or no barriers to adoption.

Social science is a particularly interesting target because the data it studies — human behavior, language, belief, social dynamics — is exactly where large language models have their deepest natural competency. GABRIEL is, in that sense, a bet on AI doing what it was always going to be best at: understanding text produced by people, at scale.