Skip to content
NewA calmer, faster ZeroHour

OpenAI: agents finish nearly half of multi-hour research tasks

Original title: Research acceleration: The jagged frontier

OpenAI News ()13 min read1
AI summary · x-ai/grok-4.7Summaries by Vani AI

Median daily agent spend hit $2,400 in October; Safety workweeks rose 77% from January to September.

OpenAI published internal measurements of how coding agents are changing its research cycle. Estimated autonomous success on two-to-four-hour tasks rose from 23% in March to 46% in October, and from 13% to 35% on four-to-eight-hour tasks. The lab says humans remain central, heavy users are pulling ahead, and it wants labs to report recursive self-improvement indicators without treating the report as a plan for full RSI.

Unconfirmed: The success rates and spend figures are OpenAI’s internal estimates, not independently verified.

Key facts
  • 90th-percentile daily usage reached $23,000 in October, up 105% from September.
  • Median researchers would trade 30% of team compute, or 40% of headcount, to keep agents for three months.
  • Success was 79% for monitoring training runs, 35% for launching them, and 25% for inference code.
  • Safety and Alignment growth accelerated after OpenAI’s response to the Hugging Face incident in July 2026.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

From the original article

13 min read at openai.com

OpenAI measures how coding agents are accelerating AI research, where their performance remains uneven, and what this signals about recursive self-improvement.

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.