OpenAI data shows AI agents reliably work only 15 minutes

OpenAI's internal estimates show AI agents can only reliably work fully autonomously for about 15 minutes, far below METR's benchmark estimate of roughly 3 hours, highlighting a gap between benchmarks and real-world tasks.

2026-10-24 ~ 2026-10-24 · 2 related posts

• OpenAI internal R&D tasks show 15-minute automation horizon vs METR's 3 hours, a 12x gap — srchvrs · 2026-10-24
• OpenAI's real-world research tasks show AI agents reliably handle only ~15 minutes — srchvrs · 2026-10-24