OpenAI's autonomous coding agents have nearly doubled their success rate on medium-length research tasks, according to a report published on the company's official website. The completion rate for tasks estimated at two to four hours climbed from 23% in March to 46% in October, while performance on four- to eight-hour tasks rose from 13% to 35% over the same period.
Despite the gains, OpenAI emphasizes that humans remain central to the research loop. At API list prices, researchers' median daily usage reached $2,400 in October, up 62% from September, with the 90th percentile hitting $23,000, a 105% increase. Work-weeks recorded in the safety and alignment field grew 77% from January to September, marking the largest absolute increase among major research areas.
The company argues that AI labs should track and publicly disclose recursive self-improvement (RSI) metrics to ensure transparency as agent capabilities advance.
OpenAI Agents Hit 46% Success Rate on Medium Research Tasks, Doubling Since March
免責事項: Phemexニュースで提供されるコンテンツは、あくまで情報提供を目的としたものであり、第三者の記事から取得した情報の正確性・完全性・信頼性について保証するものではありません。本コンテンツは金融または投資の助言を目的としたものではなく、投資に関する最終判断はご自身での調査と、信頼できる専門家への相談を踏まえて行ってください。
