Back to today’s briefing

Story perspectives

OpenAI's GDPval: AI Outperforms Humans in Routine Tasks

10/2/2025

38 5

Summary

Story summary

OpenAI released GDPval, a benchmark evaluating AI performance in 44 occupations.
GDPval compares AI models to human experts in GDP-contributing industries.
Claude Opus 4.1 leads with 47.6% win/tie; GPT-5 at 38.8%.
OpenAI says AI can be 100 times faster and cheaper, but emphasizes AI cannot fully replace human jobs.
Evaluation shows AI can handle routine tasks, freeing humans for work.

1 of 1

Story summary

OpenAI released GDPval, a benchmark evaluating AI performance in 44 occupations.
GDPval compares AI models to human experts in GDP-contributing industries.
Claude Opus 4.1 leads with 47.6% win/tie; GPT-5 at 38.8%.
OpenAI says AI can be 100 times faster and cheaper, but emphasizes AI cannot fully replace human jobs.
Evaluation shows AI can handle routine tasks, freeing humans for work.