OpenAI Researchers Use AI for Coding, Daily Costs Exceed $7,000
OpenAI has disclosed the usage scale and costs of AI programming agents among its researchers. The most frequent users are now consuming over $7,000 worth of tokens daily.
AI research firm OpenAI has revealed the extent to which its researchers are utilizing AI-powered programming agents, detailing significant daily costs for the most active users. According to a company blog post, a segment of researchers is now spending over $7,000 per day on "tokens," the fundamental units of text processed by AI models.
This technology has reportedly accelerated code writing and delivery for OpenAI's researchers over the past year, enabling more experiments and complex tasks to be assigned to AI agents. The company noted a surge in agent usage within its research department over the last three months, outpacing growth in other internal teams. As of mid-August, the median daily token cost per researcher exceeded $600, a substantial increase from $162 in July.
OpenAI clarified that these figures are estimated based on public API inference prices, representing theoretical costs at market rates rather than actual expenses. The agents are increasingly employed for troubleshooting technical issues, leading to a more than 50% reduction in daily posts on internal support channels since January.
The company aims to develop a fully automated AI researcher by March 2028, fulfilling a goal set by CEO Sam Altman for an "automated research intern" capable of completing tasks with human guidance. The broader software engineering industry is also seeing shifts, with developers delegating tasks to AI and focusing on review. However, high usage has raised concerns about costs, with some companies like Meta and Amazon reportedly ceasing internal leaderboards for AI usage to curb what is sometimes termed "tokenmaxxing."