At 23:25 UTC on September 23, OpenAI Status reported that mobile users were unable to see Work Mode or the model picker in ChatGPT. The incident notice did not describe a broad conversation outage or say that the underlying models were unavailable. The controls themselves were missing for affected users.

That distinction is easy to miss. ChatGPT can appear otherwise functional while a missing model picker prevents users from explicitly selecting the model or reasoning mode they want. If Work Mode disappears from the interface, the problem is even more straightforward: the feature has no visible entry point.

A mitigation arrived seven minutes later

OpenAI posted the initial incident update at 23:25 UTC. At 23:32, the company said it had applied a mitigation and was monitoring recovery. The public incident record did not provide a technical root cause.

Services mirroring OpenAI’s status feed subsequently recorded the incident as resolved at 23:49 UTC, with all affected services reported as recovered. That makes it a short disruption, although it would have been immediately obvious to anyone trying to use either missing control on a phone.

OpenAI has not publicly said whether the issue was tied to a specific app version, one mobile operating system, or an interface rollout. There is not enough published information to assign it a more specific cause.

ChatGPT had already seen another issue earlier that day

Several hours earlier, OpenAI had recorded a separate incident involving elevated conversation errors for Plus and Pro users. The company identified the problem at 09:13 UTC, moved to monitoring after a mitigation at 09:45 and reported full recovery at 09:54.

Two incidents appearing on the same date does not establish that they shared a cause. The earlier event involved conversation errors, while the later one concerned the visibility of mobile interface features. OpenAI has not linked the two.

A minor incident in a part of the product users increasingly rely on

The model picker has become more consequential as ChatGPT has accumulated different models and reasoning options. Losing it temporarily does not mean the underlying AI stopped responding, but it removes the control used to decide which system should answer.

Work Mode demonstrates the same dependency from another direction. A service can be operational in the backend and still be effectively unavailable to a user if its interface entry point vanishes.

For now, OpenAI’s latest publicly recorded problem is therefore a short-lived mobile interface regression: mitigation was applied quickly, recovery followed, and no technical root cause has been published.