OpenAI Says Coding Agents Now Complete 46% of 2-4 Hour Researcher Tasks, Up From 23% in March
Summary
OpenAI published 'Research acceleration: The jagged frontier' on Nov 18, reporting that coding agents now finish 46% of its 2-4 hour research tasks without human help, up from 23% in March, with 4-8 hour tasks rising from 13% to 35%. Median researcher agent spend hit roughly $2,400/day at API rates in October, a 62% jump from September, and more than 10% of researchers now run 16+ agents in parallel. Capability is jagged: model-training monitoring hits 79% while inference/serving code only 25%.
Originally reported by openai.com
Read the original article →Original headline: OpenAI Says Coding Agents Now Complete 46% of 2-4 Hour Researcher Tasks, Up From 23% in March