September 30, 2025, OpenAI released the GDPval-v0 evaluation tool and related research report, ringing the alarm bell globally about AI reshaping the job market. This systematic evaluation covering 44 professions and over 1,300 tasks across nine core industries in the U.S. shows that current top AI models have reached human expert levels in nearly half of professional tasks, with disruptive advantages in speed and cost, indicating that the acceleration phase of a global occupational structure transformation has begun.
The evaluation reveals a shocking reality: AI is approaching human professional standards
The GDPval-v0 evaluation system introduced by OpenAI is regarded by the industry as the “golden standard” for quantifying the economic impact of AI, with significance comparable to the SWE-Bench benchmark test in the programming field. The system was designed and blindly evaluated by experts with an average of 14 years of industry experience, requiring AI to generate diverse deliverables such as documents, presentations, and charts, rather than simple text-based Q&A, to fully simulate real-world work scenarios.
The evaluation results show that AI models have achieved breakthrough progress in handling professional tasks. Anthropic’s Claude Opus 4.1 topped the list with a 47.6% “win or draw” rate, particularly skilled in data visualization and document beautification, with its generated presentations rated as “superior” by experts in 49% of cases; GPT-5-high followed closely with a 38.8% rate, excelling in high-precision tasks such as legal clause interpretation; o3 high ranked third with a 34.1% rate.
More striking is the efficiency advantage of AI: cutting-edge models complete the same tasks 100 times faster than humans and at only 1/100th the cost of human experts. Taking software development as an example, the cost for AI to write basic code is $0.02 per line, while a junior programmer costs $2 per line, and AI is still 50 times faster; in government policy analysis tasks, the cost for AI to generate solutions is only $0.3 per report, far below the $30 per report of human experts.
Industry Penetration Map: No Sector is Immune
Among the nine industries covered by GDPval’s assessment, the trend of AI replacement has become clear. In the retail industry, a chain supermarket that introduced AI to optimize restocking strategies saw its inventory turnover rate increase by 15% and save over a million dollars in annual costs; in the wholesale industry, for supply chain risk assessment, AI processes 100,000 data points in just 3 minutes, while humans take 5 hours.
Professional service fields face structural impacts: In the legal industry, AI reviewing contracts is 40 times more efficient than manual work, and standardized documents can be directly generated as templates; in accounting positions, AI tools can complete report verification that would take humans 40 hours in just 3 minutes; in the medical field, although primarily focused on “augmentation,” AI can complete 90% of preliminary image screening, but the documentation work of medical secretaries has been significantly streamlined.
The substitution effect in the tech industry has already begun to manifest. Microsoft laid off 6,000 employees in May 2025, with CEO Satya Nadella admitting that AI has generated 20% to 30% of the company’s code; IBM laid off 9,000 employees around the same time, transferring some positions to low-cost regions while increasing investments in the AI field. Harvard research shows that companies adopting generative AI have seen an average reduction of 7.7% in the number of junior positions, creating a “credential-skewed technological shift.”
The substitution logic is clear: standardized tasks are hit first
The impact of AI on professions follows a clear “substitution rule,” with highly standardized, data-intensive tasks being the hardest hit. Analysis of the OpenAI GPT-3 model reveals that tax preparers have a 98% probability of being replaced, followed closely by data entry clerks, bookkeepers, and legal assistants—positions all characterized by fixed processes and clear input-output patterns.
Jack Clark, former policy director at OpenAI and co-founder of Anthropic, points out that the GDPval checklist nearly encompasses all key knowledge-intensive jobs in the modern economy, and AI is methodically infiltrating every “niche” of the economy. He emphasizes particularly, “Current models are already very close to human performance and will soon surpass humans in more tasks.”
It is worth noting that AI replacement exhibits the “middle squeeze” phenomenon. A Harvard study found that the group most severely impacted is not those with the lowest education levels, but job seekers who graduated from ordinary institutions; while demand for entry-level positions has sharply declined, senior employees have gained more promotion opportunities due to their ability to operate AI systems, and the value of high-end consulting, complex decision-making, and other positions has Conversely increased.
The transformation is upon us: How can humans break through?
In the face of the AI replacement wave, the labor market has already shown adaptive adjustments. Maimai data shows that in 2025, the number of new positions in the AI field has increased by 10 times year-on-year, with 42.66% of fresh graduate positions offering average monthly salaries of 50,000 to 80,000 yuan, indicating the rapid rise of new professions such as prompt engineering and AI system management.
Experts suggest that individuals need to build “AI immunity,” focusing on developing skills that AI cannot replace, such as emotional resonance, cross-disciplinary integration, and ethical judgment; companies should adopt the “AI-generated draft + human refinement” model, such as a design company that uses AI to generate sketches and designers to optimize creativity, shortening project cycles by 40%; at the government level, it is necessary to establish “AI replacement early warning mechanisms” and retraining funds, with Germany having achieved a 75% reemployment rate for affected workers through similar measures.
OpenAI’s evaluation report is not a declaration of the end times, but it undoubtedly reveals a certain future: as Sam Altman, the “father of ChatGPT,” said, AI is drastically shortening the 50-year job replacement cycle. In this transformation, only individuals and organizations who proactively embrace change and continuously enhance their unique value can stand firm in the new era of human-machine collaboration.
