OpenAI's researchers are using AI agents at a pace the company itself describes as exponential. According to the report “Research acceleration: The jagged frontier”, published on Wednesday (November 18), the median OpenAI researcher now consumes about $2,400 per day in agentic AI tools, and the top 10% of users reach $23,000 per day — up 62% and 105% from September, at API list prices. The figures are OpenAI's own internal measurements and have not been independently verified.
Agents now complete nearly half of research tasks
From March to October, the success rate of agents completing on their own research tasks estimated at two to four hours of human work rose from 23% to 46%; for four-to-eight-hour tasks, it rose from 13% to 35%. Performance varies widely by task type: monitoring training runs succeeds 79% of the time, answering technical questions 72%, and writing code for inference and serving stacks only 25%. OpenAI sums up the picture as “jagged” and says humans remain at the center of the research loop.
5.7 agent workdays per researcher, per day
The company also measures volume in “agent workdays”: every eight hours of agent use counts as one day. The average rose from 0.5 to 5.7 agent workdays per researcher per day — an elevenfold increase — reaching about eight workdays per day on pretraining and posttraining teams. By mid-October, more than two-thirds of researchers were hitting daily peaks of at least four agents working at the same time; more than 10% reached 16 or more.
What it is worth — and what OpenAI is asking for
In September 2026, the API-equivalent value of agent use in the research organization was about twice the researchers' non-equity compensation — excluding equity, which accounts for 70% to 80% of target pay. In a survey of 101 researchers, the median answer was that they would give up 40% of team headcount or 30% of team compute to keep access to frontier coding agents for three months. Safety and Alignment saw the largest growth in documented human effort: workweeks rose 77% from January to September. OpenAI argues that labs should be required to track and publicly report indicators of recursive self-improvement — the process by which AI accelerates the development of future AIs — and says the report should not be read as an intention to pursue full self-improvement.
Image: OpenAI logo (OpenAI, public domain — PD-textlogo, via Wikimedia Commons).
Sources
- OpenAI — Research acceleration: The jagged frontier
- Crypto Briefing — OpenAI's top researchers now burn $23,000 in AI tokens a day
- FourWeekMBA — OpenAI's Research Agents Now Cost 2x Researchers' Cash Pay
Transparency note: this article was produced with the assistance of artificial intelligence, cross-checked against public posts on X and internet sources, and reviewed before publication. Please consult the original sources to confirm the information.
By Muse