New! Check service status instantly with our Chrome extension
AI research and development platform.
There were 328 OpenAI outages since January 2025 which are summarized below, including incident details, duration, and resolution information.
Not working for you?
OpenAI's Sora API experienced a brief outage, with elevated errors affecting the Sora endpoints and causing degraded functionality for users calling those APIs. The outage began around and was resolved by , lasting about 30 minutes. The issue appeared isolated to the Sora API component.
Users attempting image generation in ChatGPT faced elevated errors beginning around , with some users unable to generate images. OpenAI identified the issue and started applying a mitigation around , and error rates began returning to normal by .
Although the core image generation service recovered, OpenAI warned of ongoing issues in FedRAMP workspaces affecting multiple features such as Codex, workspace analytics, and conversation search around ; that portion moved toward full recovery by when the FedRAMP component returned to up. Throughout the incident window, user reports indicated a broad, though not universal, impact across regions as services progressively recovered.
OpenAI faced an outage affecting image generation in ChatGPT, beginning at , with some users seeing errors or blank/incomplete image results. The issue specifically affected the ChatGPT Image Generation component, causing degraded user experiences for a portion of users worldwide. This incident occurred alongside a separate FedRAMP workspace disruption that began at and impacted multiple FedRAMP scoped OpenAI features, including Codex, workspace analytics, conversation search, searching for custom GPTs, ChatGPT user invites, and the Compliance Logs Platform download endpoint. OpenAI indicated these FedRAMP issues were known and ongoing, with core functionality restored but ongoing issues in FedRAMP environments. Recovery for FedRAMP components progressed, with an update on showing the FedRAMP issues moving toward resolution.
OpenAI observed a Windows desktop app issue affecting ChatGPT conversations, with affected users encountering problems loading chat history, missing sidebar icons, and unresponsive login controls. The problem appeared to be isolated to the ChatGPT Windows app, while users could still access ChatGPT via the web and mobile apps. The first official acknowledgment arrived at , and OpenAI advised continuing to use web or mobile as a workaround as engineers investigated. Impact began surfacing around based on user reports, and the issue persisted over the next day.
OpenAI faced degraded Codex services as usage limits depleted faster than expected, impacting Codex Web, App, Codex API, CLI, and the VS Code extension. Users reported timeouts, errors, and capacity-like messages such as 'Selected model is at capacity,' with many prompting slower responses. Impact began around and persisted until as the issue unfolded and investigations progressed.
OpenAI issued a series of status updates on the Codex-related issue, including an acknowledgment that some reports were related to abuse and fraud prevention systems incorrectly rate-limiting certain accounts, with the impact described as limited and the team continuing to monitor. By , Codex Web, App, API, CLI, and VS Code extension had been restored to normal operation. A separate item on noted issues with the ChatGPT Windows desktop app, but web and mobile access remained available as a workaround while Codex elements recovered. The broader Codex surface gradually returned to standard performance over the following hours.
Ultimately, the Codex surfaces stabilized and operations returned to typical levels. The incident coincided with a high volume of user reports indicating widespread, though not uniform, impact across Codex endpoints and regions, with performance returning to baseline as services recovered.
OpenAI began experiencing broad degraded performance across ChatGPT and Codex services starting around . An initial status update noted a separate subscription-cancellation issue affecting a small subset of users, but reports soon covered a wide range of features. Users globally reported timeouts, slow responses, and disruptions across major components such as Conversations, Login, and GPTs, with related services like Image Generation, Voice mode, and File uploads affected. Codex usage limits were reported as depleting faster than expected, and some accounts were rate-limited. The Windows desktop app also saw issues on Windows, including UI elements and login controls not responding in places. The situation persisted for several days as OpenAI investigated and issued updates, with services gradually returning to normal by .
OpenAI Codex services began a broad disruption on , with users reporting access-token failures across Codex components including the API, CLI, VS Code extension, Codex Web, and the App. Over the next hours, FedRAMP workspaces and API orgs also showed degraded performance, prompting mitigation and ongoing monitoring by . The incident signaled a multi-surface disruption that affected authentication and core Codex access for many workflows.
By , reports indicated degraded performance for FedRAMP workspaces and API orgs, with some components moving back to normal as mitigations took effect but investigations continued. Later, around and again at , OpenAI flagged Codex usage limits depleting faster than expected, with follow-up noting that abuse-prevention systems were incorrectly rate-limiting certain accounts. The issue produced recurrent capacity and rate-limit errors across Codex Web, App, API, CLI, and the VS Code extension.
As of , user reports indicated the disruption remained widespread with global distribution, including slow responses, timeouts, and stream/disconnect errors across Codex workloads. OpenAI continued investigations and monitoring of mitigations, with no public resolution time provided.
The incident began at when gpt-4o-mini reported a high error rate, triggering broad degradation across OpenAI APIs. Users experienced timeouts, failed requests, and slower responses across a wide range of features, with affected components including APIs / Batch, / Files, / Login, / Sora, / Moderations, / Realtime, / Audio, / Embeddings, / Images, / Chat Completions, / Responses, and / Fine-tuning. The disruption spanned multiple days and was observed globally across chat, coding, and media-related workflows.
From , OpenAI users faced elevated errors and degraded performance for ChatGPT file uploads and downloads, with reports spanning worldwide. The issue broadened to FedRAMP workspaces and API orgs and affected a wide swath of APIs, including ChatGPT / File uploads and endpoints such as Batch, Files, Login, Sora, Moderations, Realtime, Audio, Embeddings, Images, Chat Completions, Responses, and Fine-tuning. A high error rate was also observed for gpt-4o-mini. By mitigations were in place and most components began to recover, though some services remained degraded for a time. The incident persisted for several days with user reports continuing globally, and status returned to up by as OpenAI completed the recovery.
OpenAI's platform experienced broad degraded performance and elevated errors across multiple services, affecting both ChatGPT features and developer APIs. Users reported timeouts, slow responses, and failures with tasks such as file uploads and logins, with issues spanning FedRAMP workspaces, API orgs, and a wide range of API endpoints. The impact extended to Codex-related flows and the gpt-4o-mini tier, leading to intermittent reliability challenges for both consumer and developer use cases. Impact began around and many users worldwide encountered degraded experiences over several days.
Looking for more OpenAI outage stats?
Get real-time insights and historical data on OpenAI outages.
OpenAI outage statsNo credit card required
