GDPval
OpenAI benchmark of real-world, economically valuable knowledge-work tasks across 44 occupations; a 220-task gold subset with files and rubrics is open on Hugging Face.
Tags
Description
The full set has 1,320 tasks built from real work products by professionals averaging 14 years of experience, spanning 44 occupations in the top 9 sectors of U.S. GDP. Each task has a prompt plus reference files (spreadsheets, documents, etc.); the open gold subset (openai/gdpval, v2 with rubrics and gold deliverables) has 220 tasks. OpenAI offers an experimental automated grader at evals.openai.com; human expert comparison remains the recommended grading method. Some tasks contain NSFW or political content. The Hugging Face dataset card does not state a license.
Related
- TheAgentCompany: Benchmark of 175 professional tasks in a simulated software company (self-hosted GitLab, Plane, ownCloud, RocketChat) for agents that browse, code, run programs and talk to coworkers.
- Humanity's Last Exam: Multi-modal benchmark of 2,500 expert-written questions at the frontier of human knowledge, from the Center for AI Safety and Scale AI.
- METR Task-Completion Time Horizons: METR's up-to-date measurements of how long tasks frontier AI models can complete autonomously (time horizons).
Sources
- https://huggingface.co/datasets/openai/gdpval
- https://arxiv.org/abs/2510.04374
- https://openai.com/index/gdpval/
Machine-readable
- JSON
- Markdown
- Source file on GitHub (edit via pull request)