OpenAI has acknowledged a disruption affecting ChatGPT and Codex, with users experiencing persistent errors for much of the day. The company's official system status page describes a partial outage that has lasted at least four hours, leaving many unable to access the services as usual.
OpenAI described the incident as “elevated error rates.” The company said it began investigating the problem several hours ago and implemented a mitigation roughly two hours into the outage. It is monitoring the recovery, but the issue remains unresolved for some users. That pattern, applying a mitigation and then monitoring recovery, also appeared in a separate OpenAI Status incident on October 20, when the company reported degraded performance in ChatGPT, Codex and APIs, including Codex Web, API and CLI, before marking those services recovered.
OpenAI also documented a separate partial outage on October 20 involving authentication errors for some users of its API, ChatGPT and Codex. The company said the incident lasted from 9:35 to 10:35 a.m. PDT, affected some users near Illinois and was resolved after it rolled back a configuration change. The incident shows how a change affecting shared access systems can disrupt multiple products, though OpenAI has not identified that configuration issue as the cause of the outage described here.
External monitoring platforms such as DownDetector have shown a decline in user reports after an initial surge, suggesting some improvement. But, ongoing complaints on social media, particularly on X, indicate that the disruption continues to affect users. In a public issue opened in the openai/codex GitHub repository, users also described receiving “Selected model is at capacity” errors across Codex, ChatGPT and ChatGPT Work from October 22. That user report said the errors persisted across several models and affected a ChatGPT Pro account with unused quota; it does not establish the cause of the present disruption.
Capacity problems have featured in earlier Codex incidents as well. An analysis of OpenAI status records identified incidents on June 16, July 9 and July 17, 2026, lasting 196, 58 and 203 minutes, respectively, involving model-capacity or overload errors. Those episodes provide context for the service’s reliability history, but do not establish that capacity is responsible for this outage. OpenAI’s help documentation also confirms that Codex can be used with a ChatGPT plan and that model availability for ChatGPT-account sign-ins can change, showing the account and service relationship between the products.
ChatGPT and Codex support a range of applications, from conversational AI to code generation, and their reliability matters to individual users and businesses. Interruptions can disrupt workflows and delay projects, leaving people who rely on the tools in their daily operations uncertain. OpenAI's response and updates on its status page have provided some reassurance, but the outage remains unresolved for a portion of users.
Disruptions affecting major AI services such as ChatGPT and Codex can have wide-reaching effects because the tools are integrated into professional and personal routines. OpenAI Status recorded other ChatGPT- and Codex-related incidents in October, while independent outage tracking has also counted incidents across AI providers; these records vary in scope and severity. As reliance on AI tools grows, users and industry observers are likely to scrutinize how providers communicate incidents and restore service. OpenAI has committed to updating users when more information becomes available or the issue is fully resolved.
