OpenAI says GPT-5 stacks up to humans in a wide range of jobs

OpenAI has introduced GDPval, a new benchmark designed to assess AI models' performance against human professionals across 44 occupations in nine key industries, including finance and healthcare. The initial results show OpenAI's GPT-5-high matching or surpassing human experts in 40.6% of tasks, while Anthropic's Claude Opus 4.1 achieved 49%, representing a significant leap from GPT-4o's 13.7% just 15 months prior. Although GDPval-v0 currently focuses on report generation, these findings highlight the accelerating pace of AI development towards expert-level quality in specific professional tasks, suggesting potential for substantial productivity gains and a shift towards higher-value human work as AI capabilities advance.

Analysis

OpenAI has introduced a new benchmark, GDPval, to quantify AI model performance against human experts across 44 occupations in key economic sectors. The initial results indicate that foundation models are rapidly approaching expert-level quality in specific tasks, with Anthropic's Claude Opus 4.1 being rated as on par with or better than human professionals in 49% of tested scenarios, and OpenAI's own GPT-5-high achieving a 40.6% rating. This marks a substantial capability leap from OpenAI's GPT-4o, which scored only 13.7% just 15 months prior, underscoring the accelerating pace of AI development. While OpenAI notes the current benchmark is limited to report generation and suggests Claude's higher score may be partly due to superior presentation, the trend points toward a significant potential for productivity augmentation in white-collar jobs. The creation of economically-focused benchmarks like GDPval represents a maturing evaluation landscape for AI, shifting focus from academic tests to real-world, value-generating applications.

Key Decisions for Investors

• The rapid performance improvement from GPT-4o to GPT-5 validates the long-term investment thesis in foundational AI models and the broader technology sector, suggesting that the pace of innovation remains high.
• Investors should monitor companies in sectors like finance and healthcare for early adoption of these advanced AI tools, as they are positioned to realize significant productivity gains and operational efficiencies.
• The competitive results from Anthropic's model indicate a dynamic market where technological leadership is not guaranteed; therefore, tracking the evolution of real-world benchmarks like GDPval will be crucial for assessing the competitive positioning of key AI developers.
• Given the current test's limitation to report generation, investors should view this as a signal for AI's role in task augmentation rather than imminent, large-scale job replacement, tempering expectations for immediate labor market disruption.