Story perspectives
OpenAI's GDPval Benchmarks AI Against Human Professionals
9/26/2025
41 9
1 of 1
Story summary
- OpenAI launched GDPval, a benchmark to compare AI with human professionals across 44 occupations in nine sectors, including healthcare and finance.
- GPT-5 achieved a win rate of 40.6%, while Anthropic's Claude Opus 4.1 scored 49%.
- OpenAI notes GDPval measures only task-level performance, not full job functions.
- Future tests will cover more industries and interactive processes to gauge AI impact.
