Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models

GPT-5.5 completes tasks that take humans three minutes with 50% reliability. No-CoT task completion time horizons have doubled annually since 2019. The gap between with-CoT and no-CoT growth rates widened after GPT-4, with no-CoT growing at half the rate of with-CoT. Frontier models now exceed human task duration in extended reasoning without chain-of-thought.