--- title: OpenAI's fourth outage in four days hits ChatGPT, Codex,... description: OpenAI experienced its fourth service disruption in four days, with simultaneous outages affecting its APIs, ChatGPT, and Codex. Users encountered 503 errors... canonical: https://daily.dev/posts/openai-s-fourth-outage-in-four-days-hits-chatgpt-codex-and-apis-simultaneously-vwsr66d95 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: OpenAI's fourth outage in four days hits ChatGPT, Codex, and APIs simultaneously | daily.dev og:description: OpenAI experienced its fourth service disruption in four days, with simultaneous outages affecting its APIs, ChatGPT, and Codex. Users encountered 503 errors... og:url: https://daily.dev/posts/openai-s-fourth-outage-in-four-days-hits-chatgpt-codex-and-apis-simultaneously-vwsr66d95 og:image: https://api.daily.dev/og/posts/vWSR66D95.png og:image:alt: OpenAI's fourth outage in four days hits ChatGPT, Codex, and APIs simultaneously og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # OpenAI's fourth outage in four days hits ChatGPT, Codex, and APIs simultaneously **[Collections](https://daily.dev/sources/collections)** · 5 min read · 1 upvotes · 0 comments ## Summary OpenAI experienced its fourth service disruption in four days, with simultaneous outages affecting its APIs, ChatGPT, and Codex. Users encountered 503 errors blocking requests entirely, with mitigation applied within an hour. OpenAI's status page shows roughly 166 incidents over the past nine months — about 18 per month. The reliability concern is amplified by OpenAI's 10 million weekly agentic users and the recent launch of ChatGPT Work, where mid-task outages can cascade into failed workflows and broken pipelines. Anthropic's Claude faced a similar outage in June, raising broader questions about whether AI infrastructure is ready for enterprise-grade reliability demands. ## Content ## What happened OpenAI hit another service disruption over the weekend — its fourth in four days — with elevated error rates across its APIs, ChatGPT, and Codex simultaneously. Users saw 503 errors labeled `biscuit_baker_service_me_circuit_open`, blocking requests from reaching OpenAI's servers. The company applied a mitigation within roughly an hour. Afterward, OpenAI's Tibo confirmed that usage limits had been reset for all paid users of Codex and ChatGPT Work. ## The broader reliability picture This isn't an isolated incident. OpenAI's status page has logged around 166 incidents over nine months — about 18 per month. Anthropic's Claude suffered a similar outage in June. The pattern raises a real question about whether current AI infrastructure can meet enterprise-grade reliability expectations. The stakes are higher now than they were a year ago. OpenAI has 10 million weekly agentic users, and ChatGPT Work launched recently. A 45-minute chatbot outage costs roughly 45 minutes. A 45-minute coding-agent outage can destroy several hours of effective work — the failed job may have already accumulated repository context, test results, partially completed patches, and tool state. The correct compensation unit isn't minutes unavailable; it's durable work lost. ## The reset question A global hard reset is the most visible response, but it's not obviously the best one. A few problems worth noting: **It rewards users unevenly.** A user with 5% of their allowance remaining gets enormous value from a reset. A user with 95% remaining gets almost nothing — and loses the strategic value of quota they deliberately conserved. **It can trigger a second outage.** When millions of users receive fresh quota simultaneously, many immediately restart large jobs. That creates a thundering herd of retries, synchronized repository cloning, and rapid queue growth — additional pressure on infrastructure that just recovered. **It obscures the actual accounting.** A blanket reset doesn't reveal whether failed tasks were charged, whether retries were double-counted, or whether partial work can be resumed. A banked reset — where users receive a reset they can activate when needed — would address most of these problems. It preserves unused quota, avoids synchronized demand, and provides equal option value to all affected users. OpenAI already introduced banked Codex resets for eligible Plus and Pro users in June 2026, so the infrastructure exists. ## What better compensation infrastructure would look like The cleanest solution is transactional quota metering: quota is reserved at task start, accumulates provisionally during execution, and is permanently charged only when meaningful work is durably committed — a checkpoint saved, a patch published, test results preserved. If a platform-side failure occurs before durable delivery, the reserved usage is automatically released. No durable output, no permanent charge. Short of that, incident-aware automatic refunds would help. Every failed task carries a task ID, incident ID, usage consumed, and failure ownership classification. When OpenAI confirms a platform incident, an automated reconciliation job restores affected usage without requiring support tickets. Users see something like: "2.4% of your weekly Codex allowance was restored because task X failed during incident Y." Durable resumability matters too. After recovery, users should be able to continue from the final valid checkpoint rather than restarting the complete task from scratch. ## The competitive angle OpenAI has issued several Codex and ChatGPT Work resets around the GPT-5.6 Sol rollout and user milestones — the most recent major one was tied to the 10-million-user milestone on July 21, just days before this incident. Resets have become unusually frequent. Anthropically, Anthropic did not attach a fresh global reset to the Opus 5 launch on July 24, though Claude Code's weekly limits were already temporarily 50% higher through August 19 for Pro, Max, Team, and eligible Enterprise users. Anthropic did reset five-hour and weekly limits on Sol's launch day — interesting timing, though timing alone doesn't prove it was a competitive response. The frontier-model competition is quietly expanding beyond benchmark scores and subscription prices. What increasingly matters is how much productive agent work fits inside a subscription, whether interrupted tasks consume quota, whether long jobs can resume, and how quickly failed usage is restored. A model can be substantially more capable and still deliver less practical value if its jobs frequently stall or consume scarce quota without producing a durable result. Repeated surprise resets also create a subtler problem: when users expect frequent unannounced resets, rational behavior shifts. They stop conserving weekly usage, launch speculative tasks near the quota boundary, and treat the nominal quota as a soft constraint. That damages the provider's demand telemetry — it becomes hard to tell whether usage reflects genuine workload or strategic "burn it before the next reset" behavior. A formal, predictable compensation policy — published rules about what triggers automatic refunds versus banked resets versus promotional resets — would reduce that distortion and make resets feel like infrastructure rather than personality-driven social announcements. --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#openai](https://daily.dev/tags/openai), [#chatgpt](https://daily.dev/tags/chatgpt) [View this post on daily.dev](https://daily.dev/posts/openai-s-fourth-outage-in-four-days-hits-chatgpt-codex-and-apis-simultaneously-vwsr66d95) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"OpenAI's fourth outage in four days hits ChatGPT, Codex, and APIs simultaneously","url":"https://daily.dev/posts/openai-s-fourth-outage-in-four-days-hits-chatgpt-codex-and-apis-simultaneously-vwsr66d95","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/openai-s-fourth-outage-in-four-days-hits-chatgpt-codex-and-apis-simultaneously-vwsr66d95"},"datePublished":"2026-07-25T11:51:37.225Z","dateModified":"2026-07-28T03:10:21.772Z","description":"OpenAI experienced its fourth service disruption in four days, with simultaneous outages affecting its APIs, ChatGPT, and Codex. Users encountered 503 errors...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/407136d433d7cb4c2755bdfb34dbd3c9?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/407136d433d7cb4c2755bdfb34dbd3c9?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/openai-s-fourth-outage-in-four-days-hits-chatgpt-codex-and-apis-simultaneously-vwsr66d95","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,openai,chatgpt","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"OpenAI's fourth outage in four days hits ChatGPT, Codex, and APIs simultaneously"}]} ```