OpenAI Codex lead Tibo confirmed today that the team has identified and fixed two bugs that were responsible for GPT-5.5’s degraded performance in the terminal-based coding agent Codex over the past two days. The system is now fully restored. To make it up to developers, OpenAI has reset all users’ usage limits.
https://twitter.com/thsottiaux/status/2055540700092211404
Over the last 48 hours, some users had been reporting that GPT-5.5 in Codex seemed less capable. Early in the investigation, OpenAI couldn’t pinpoint the cause. But this afternoon, monitoring metrics show everything is stable again.