OpenAI Codex Shell execution latency regressed 8-11x on Windows

maikolb · ghdev · 2026-09-01

A user reported a severe performance regression in OpenAI Codex CLI on Windows, with shell command execution latency increasing from 1.7s to 18.4s (median) between versions 0.146.0 and 0.151.0-alpha. The issue spans 10 months of rollout data. Doctor reports indicate a sandbox provisioning failure where the system forces an elevated backend despite the user's configuration, leading to the latency spike.

Original post →

More from Apps

Apps channel →