OpenAI Experiences 4-Hour Outage Affecting ChatGPT, Codex, and API Platforms

According to monitoring by Beating, starting at 01:35 Beijing time on April 21, OpenAI’s status page began investigating an outage affecting ChatGPT, Codex, and the API platform. Users reported widespread loading issues, and the official classification was a partial outage. The affected components included 12 for ChatGPT and 1 for Codex. The outage lasted approximately 4 hours and 13 minutes. The official updates on mitigation were released at 03:48, 04:43, and 05:17 (Beijing time), with full restoration announced at 05:48. During the same time frame, the error rate for ChatGPT conversations in Europe also increased, with mitigation starting at 03:50 and 04:49. Earlier, starting from 18:13 on April 20, ChatGPT Business experienced an anomaly lasting up to one hour during upgrades or new seat additions, with monitoring for recovery beginning at 20:43. Less than two hours after the outage ended, Codex experienced a brief failure around 7:00 AM on April 21. Codex engineering lead Thomas Sottiaux later reported on X that the incident lasted about 10 minutes. He mentioned that Codex contributed to the repair process, but the final key insights came from humans. A notification will be sent to affected users regarding rate limit resets as compensation.