Per OpenAI's status page, users experienced "elevated error rates for GPT 5.5 in Codex" and saw "model at capacity" messages earlier on Jun 16, 2026. Business Insider reports the first dashboard error appeared at 6:32 a.m. Eastern Time, and OpenAI's status indicator showed recovery at 9:48 a.m. Eastern Time, with the status page saying "All impacted services have now fully recovered." Business Insider also cites a post on X from Codex lead Thibault Sottiaux acknowledging high error rates and noting mitigation work. Community posts on OpenAI's developer forum and third-party outage aggregators recorded user reports and a parallel incident history for Codex and ChatGPT features. The incident is resolved; practitioners should note the timeline and monitor availability for coding workflows that depend on GPT 5.5.
What happened
Per OpenAI's status page, the company reported "Elevated error rates for GPT 5.5 in Codex" during a Jun 16, 2026 incident (OpenAI Status). Business Insider reports the first dashboard error was recorded at 6:32 a.m. Eastern Time, and that Codex users encountered "model at capacity" and other high error rates on social media (Business Insider). Business Insider also quotes Codex lead Thibault Sottiaux on X: "We are aware that some Codex users are experiencing high error rates with 'model at capacity' and are working to bring things back to being stable." OpenAI's status indicator showed recovery at 9:48 a.m. Eastern Time, and the status page stated, "All impacted services have now fully recovered." Community posts on OpenAI's developer forum and outage aggregators captured user reports and follow-up discussion (OpenAI community; IsDown.app).
Editorial analysis - technical context
Incident descriptions and the "model at capacity" messages correspond to classical capacity and rate-related failure modes in hosted model APIs. Industry-pattern observations: large transformer-based models exposed via shared API endpoints commonly surface "at capacity" errors when demand spikes or when capacity provisioning lags. Observers often see mitigations such as routing traffic to older model versions, throttling, or temporary rate limiting while operators restore full capacity.
Context and significance
Codex is a developer-facing code generation and automation tool built on frontier models; Business Insider frames it as a competitive product in a market that includes Anthropic's Code offerings. For practitioners, outages that last multiple hours can meaningfully disrupt CI pipelines, editor integrations, and automated code tasks that rely on stable API access. Public status-page transparency and a named engineering lead posting on X reduce uncertainty for integrators, but community complaints on the developer forum show customer impact and frustration.
What to watch
For observers and engineering teams integrating Codex, track these indicators: the OpenAI status page for recurrence or post-incident notes, community forum threads for reported edge cases, and any SDK or client updates that change retry or backoff behavior. Industry-pattern observations: teams using hosted generative models typically implement multi-model fallbacks, exponential backoff, and circuit-breakers to reduce disruption when "model at capacity" errors appear.
Bottom line
This incident was resolved within several hours, with OpenAI's status page reporting full recovery. Developers relying on GPT 5.5 in Codex should treat availability risk as an operational consideration and monitor OpenAI status updates and community reports for any follow-up details.
Scoring Rationale #
A resolved outage in a widely used developer model is notable for practitioners because it disrupts coding workflows and highlights availability risk for hosted APIs. The story is operationally relevant but not a structural industry shift.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.