OpenAI has pulled back the curtain on how much its researchers are spending AI coding tools.
In a blog post on Sunday, the startup said the most prolific users of AI coding tools among its researchers now consume more than $7,000 worth of tokens — the small chunks of text that an AI model reads and produces — per day.
The technology has substantially changed the way OpenAI researchers work over the past year, the company said, with staff shipping code faster, running more experiments and delegating increasingly high-level tasks to coding agents.
OpenAI said the use of AI coding agents among its researchers has skyrocketed over the past three months, outpacing growth among other OpenAI teams.
As of mid-August, the average researcher at the company’s research organization was using more than $600 worth of tokens per day, up from $162 in July, according to OpenAI data, with 90th percentile users spending more than $7,000 per day. The numbers are based on inferences based on API prices, meaning they are estimates of what they would cost rather than the dollar amount OpenAI actually paid.
OpenAI also said its agents were increasingly being used to troubleshoot technical issues, so much so that the number of daily messages in a human-staffed internal technical support channel fell by more than half since January.
The company said it had achieved the goal – set last year by CEO Sam Altman – of becoming a ‘automated research intern” that can perform tasks under human direction, and was on track to create a fully automated AI researcher by March 2028.
The Great Reset of Coding
The growing popularity and proficiency of coding tools has had a radical impact transformed software engineering the past year.
Many developers now delegate their coding tasks to AI agents and review their output. Some say this makes them more productive, but others are less on board.
It has also proven expensive for some companies.
Earlier this year, companies encouraged employees to experiment with AI, including rankings to follow the main users. That led some employees to engage in “tokenmaxxing,” the idea of using AI as much as possible regardless of output.
By the summer, however, the rising costs of top AI models began fueling concerns that token maxxing was not increasing productivity enough to justify the expenditure.
Companies like Meta and Amazon have closed internal token rankingswith Amazon senior vice president Dave Treadwell warning staff not to “use AI just for the sake of using AI.”
OpenAI has previously embraced high token usage. Altman said in June that the company’s top token user is burning out 100 billion tokens a month.
OpenClaw creator Peter Steinbergerwho was hired by OpenAI in February, posted a screenshot of a $1.3 million monthly token bill in May — though he, like other employees, doesn’t have to pay the bill.
