OpenAI’s GDPval benchmark demonstrates AI’s economic breakthrough
OpenAI’s GDPval benchmark, released in September 2025, marks the first systematic demonstration of AI systems matching human expert performance across economically valuable professional tasks. Claude Opus 4.1 achieved a 49% win rate against human experts, while GPT-5 reached 40.6%—representing more than a tripling of performance from GPT-4o’s 13.7% just 15 months earlier. Most significantly, AI…