OpenAI says coding agents finish 46 percent of mid-length research tasks

On Nov. 18, 2026 OpenAI published an internal update on research acceleration. It said estimated success without human intervention rose from 23 percent in March to 46 percent in October for tasks it estimates take a person two to four hours, and from 13 percent to 35 percent for four-to-eight-hour tasks. For Aug. 1 through Oct. 18 it put success at 79 percent for monitoring training runs, 35 percent for launching them, and 25 percent for inference and serving code. OpenAI said the rates can also reflect which tasks researchers delegate, and that the report is not a plan to pursue full recursive self-improvement.

/ Sources

/ About this story

Compiled by Venture Atlas from the sources above, using automated AI-assisted research. This is a summary of reporting published elsewhere, not original reporting - follow the source links for the full account. See our editorial standards.

Something wrong here? Email apostructura@gmail.com and we will fix it.

/ Related