OpenAI Outage History

OpenAI API users experienced increased latency across the Realtime, Embeddings, and Completions (legacy) endpoints, causing slower responses for chat interactions, text generation, and embeddings tasks. The incident began at Aug 21, 2025 7:16 PM UTC as the status page moved to a degraded state and affected three components. The outage lasted approximately 3 hours and 55 minutes, with delays and timeouts reported by users during this window.

OpenAI issued three status updates during the incident. The first at Aug 21, 2025 7:16 PM UTC noted increased latency across the APIs; a follow-up at Aug 21, 2025 8:16 PM UTC indicated recovery on the chat-completions and text-embedding endpoints while realtime latency remained under investigation; the final update at Aug 21, 2025 11:11 PM UTC stated that the Realtime API had returned to normal. The impact appeared broad, affecting users globally with varying experience in chat, generation, and embedding workflows.

OpenAI Sora services experienced a short outage with elevated error rates and latency across five Sora components: Login, Image Generation, Video generation, Video viewing, and Feed. The incident began at Aug 20, 2025 6:21 PM UTC and was resolved by Aug 20, 2025 6:51 PM UTC. The provider posted 2 status updates during the incident and noted that mitigation steps were being implemented.

Impact was felt by users who could not sign in, generate content, or access prompts and feeds. Reports described errors loading pages, 404 messages, and prompts not loading or uploading, with issues appearing across multiple regions during the incident window. The volume and spread of user reports indicated a widespread disruption during this period.

OpenAI's Sora team implemented mitigation steps per the status updates, and services returned to normal after about 30 minutes. The affected components were restored to normal operation as the incident was resolved.

OpenAI experienced degraded performance for ChatGPT Conversations beginning at Aug 20, 2025 3:16 PM UTC. Users across regions reported errors, unresponsive chats, and loading issues, particularly when starting or continuing conversations. The issue affected only the ChatGPT Conversations component and did not indicate a full OpenAI service outage. The incident prompted official status updates as the situation developed. The outage lasted until Aug 20, 2025 5:41 PM UTC when the component appeared to recover.

OpenAI reported a subscription loading issue that impacted multiple entry points for users, including ChatGPT Web, ChatGPT Desktop (Windows and macOS), and the Login flow. Users experienced failures loading conversations, signing in, or launching the app for several hours. The outage began on Aug 20, 2025 10:06 AM UTC and was resolved by Aug 20, 2025 3:11 PM UTC, with the status page documenting a mitigation and subsequent recovery across affected components. OpenAI issued three official status updates during the incident; macOS Desktop moved to Up in the mid-session update, and by the final update Windows Desktop, Login, and Web had returned to normal operation. The incident generated a high volume of user reports from multiple regions, indicating broad geographic reach and widespread disruption across services.

OpenAI's ChatGPT services experienced an elevated error rate and degraded performance, impacting a broad set of features across desktop, web, and mobile experiences as well as GPT-powered tools. Users reported HTTP 500 errors, unresponsive chat histories, and sign-in issues, with activity observed across multiple regions. The outage began at 6:56 AM UTC and was resolved by 8:01 AM UTC.

OpenAI experienced elevated errors affecting file uploads to vector stores within the API / Files component. Users attempting to upload files for vector storage reported delays and failures. The issue was first reflected on OpenAI's status page at Aug 19, 2025 4:46 PM UTC when the service was listed as degraded and an investigation began. The incident persisted for about 3 hours and 31 minutes and was resolved by Aug 19, 2025 8:17 PM UTC. OpenAI issued two status updates during the incident describing the investigation and the recovery.

User reports originated from multiple regions, including the United Kingdom and Indonesia, with comments describing slow image generation and upload problems tied to vector storage. The event did not indicate a full service outage beyond the file upload path to vector stores. There was no Early Warning Signal. The status returned to normal by the end of the window.

OpenAI reported a degraded condition affecting GPTs with custom actions that require confirmation, causing those GPTs to remain in a processing state after permission was granted. The issue was limited to the ChatGPT / GPTs component and specifically to the GPTs with custom actions feature, impacting users attempting to use actions that require confirmation.

The incident began at Aug 18, 2025 5:36 PM UTC when the status page transitioned to a degraded state for ChatGPT / GPTs. The issue persisted for about 3 hours and 40 minutes and was resolved by Aug 18, 2025 9:16 PM UTC, with two status updates posted during the incident.

An outage affected image loading in OpenAI's ChatGPT across mobile and desktop platforms. Images failed to render for users on ChatGPT Mobile iOS, ChatGPT Mobile Android, ChatGPT Desktop Windows, and ChatGPT Desktop macOS. The incident started at Aug 17, 2025 3:06 PM UTC when the status page moved from up to warning and continued with investigations and mitigation until it was resolved at Aug 17, 2025 6:06 PM UTC. The provider issued 3 status updates during the incident, documenting investigation and mitigation efforts. The issue impacted image display during chat interactions for several hours, with affected services returning to normal after the resolution.

OpenAI's ChatGPT Agent experienced elevated error rates starting at 10:51 PM UTC, resulting in degraded performance for some users. The issue affected the ChatGPT / Agent component, causing errors and blocked prompts for a portion of users over several hours. The outage lasted about 4 hours and 11 minutes, with the status page moving from up to warn and back to up as mitigation was implemented. The provider issued 2 status updates during the incident, describing mitigation progress and current status. A single user report from Waterloo, Ontario, Canada noted a resource not found error, illustrating the symptoms observed by affected users. By 3:02 AM UTC, OpenAI indicated recovery and the ChatGPT Agent returned to normal operation.

OpenAI reported elevated error rates affecting the ChatGPT / Agent component. The incident began at 8:46 PM UTC and was resolved by 10:41 PM UTC, lasting approximately 1 hour and 55 minutes. OpenAI issued two status updates during the incident as they worked to implement a mitigation.