GDPval: models approach experts on real occupational tasks

OpenAI's GDPval finds the best models' work rated as good as or better than industry experts' in nearly half of tasks across 44 occupations.

What happened and why it matters

GDPval tasks were drawn from the work of professionals with ~14 years of experience. Claude Opus 4.1 was rated as good as or better than experts in ~48% of comparisons, and models were ~100x faster and cheaper on the raw generation step.

Sources

• Measuring the performance of our models on real-world tasks ↗ OpenAI · 25 Sep 2025