Logins Broke On Thursday. Pro Users Were Already Getting 429s.
OpenAI spent Thursday morning investigating ChatGPT app login and signup failures, then said a mitigation was in place. Paying Pro users were still reporting 429 errors with quota left.
An out of order sign, shown for illustration. Photo: Ville Säävuori, CC BY-SA 2.0, via Wikimedia Commons.
Thursday morning OpenAI opened a new status incident, Widespread feature degradation, covering logins and signups in the ChatGPT mobile apps. By 10:13 AM Pacific the company said a mitigation was in place and it was monitoring recovery. That did not stop Pro subscribers from reporting HTTP 429 errors in Codex with weekly allowance still left.
What the status page said
At 9:29 AM Pacific the first update read: "We are investigating an issue affecting logins and signups in the ChatGPT iOS app. We will provide updates as we learn more."
Six minutes later, at 9:35 AM, the incident widened. The update said: "We are investigating widespread feature degradation, including issues affecting logins and signups in the ChatGPT iOS/Android apps. We will provide updates as we learn more."
At 10:13 AM the status changed to Monitoring: "We have applied the mitigation and are monitoring the recovery." The page listed no components as affected. OpenAI did not give a cause.
Pro users were already hitting 429s
A day earlier, a ChatGPT Pro subscriber posting on GitHub as kocian55-hash opened issue 54004. The subscription is listed as ChatGPT Pro 200, 229 euros a month. The exact error in the report is "exceeded retry limit, last status: 429 Too Many Requests." The models named are GPT 6 Astra and GPT 6.1 Sol. The usage screenshot in the report shows 80 percent of the weekly allowance remaining and 60,110 credits available. The author notes those figures are from the screenshot and do not necessarily match the allowance at the moment of each error.
OpenAI Support told that user: "Escalated to a support specialist; You can expect a response in the coming days." The same author added that 1.5 times speed was enabled and Codex was still extremely slow, including sessions that fail with the same 429 line. The feedback id on the report is no-active-thread-01a13ae2-7e25-71a2-8e53-3a92115e5a96.
On October 15 another Pro subscriber, khuyendev, wrote that a regression started around October 11, 2026, while using the official Codex app primarily with GPT 6.1 Sol. The comment says requests are being rejected with "Too Many Requests" or model capacity errors even when the subscription usage allowance has not been exhausted. It also says about 20 concurrent GPT 6.1 Sol Fast Mode requests used to run reliably before October 11, and that running multiple requests has been unreliable since then.
Two different failures, same morning
The login incident and the 429 reports are not the same outage. One is a status page event about getting into the ChatGPT iOS and Android apps. The other is paying users who are already signed in and still cannot finish a Codex request. OpenAI has not said the mitigation covers Codex capacity errors.
We will update this page if the status page moves again or if OpenAI explains either failure.
Sources
- OpenAI status: Widespread feature degradation (October 15, 2026)
- GitHub openai/codex issue 54004, opened October 14, 2026, comment October 15, 2026