OpenAI Confirms Mobile App Bug Hiding Work Mode and Model Picker as 'Codex Down' Searches Spike
OpenAI says a mobile display issue is preventing some users from seeing Work Mode and the model picker, driving a surge in 'Codex down' searches even as the service itself runs.
OpenAI has acknowledged that some users on its mobile app cannot see Work Mode or the model picker, a display problem that surfaced this week and quickly became a trending complaint. The company confirmed the issue on its status channels, framing it as a visibility bug rather than an outage of the underlying models. Search interest in "codex down" climbed past 500 queries as developers went looking for an explanation.
The distinction matters because Codex, OpenAI's coding agent, has become a daily dependency for a large pool of engineers since it was pushed into the mainstream through ChatGPT and the API. When the picker fails to render, users lose the ability to switch between models or enter the work-oriented interface, making the app appear broken even if requests would otherwise succeed. Reports point to the mobile clients specifically, leaving the desktop web experience apparently unaffected.
This is not the first time OpenAI's interface layer has lagged behind its model releases. The company has shipped a steady cadence of new reasoning and coding models over the past year, and each addition has to be wired into every client, from iOS and Android to the web app and third-party integrations. Configuration screens and mode selectors are usually the last pieces to catch up, which is where users tend to notice breakage first.
For developers, the practical fallout is a loss of control rather than lost compute. A missing model picker means teams cannot pin a specific version for reproducible work, and a hidden Work Mode removes the entry point for agentic coding tasks. Some users took to social channels to ask whether Codex had gone offline entirely, a sign of how thin the line is between a UI glitch and a perceived service failure.
OpenAI has not given a timeline for the fix or said how many accounts are affected. The open question is whether this is a client-side rendering bug that a patch resolves in days or a deeper issue with how the app reads feature flags from the server. Until the company ships a fix, mobile users who need Work Mode or a specific model should treat the desktop client as the reliable path, and watch OpenAI's status page for the all-clear.