OpenAI applies mitigation and monitors recovery after elevated API error rates

OpenAI said it had applied a mitigation and was monitoring the recovery on September 17, 2026, after identifying elevated error rates across its API models in an incident on its status page that listed 12 affected components of the company’s API platform.

Incident Disclosure and Response

The incident record, titled “We are seeing elevated error rates across API models,” carries a degraded-performance impact designation. In its first update, timestamped 4:20 PM ET on September 17, 2026, OpenAI said it had identified that users were experiencing elevated errors for the impacted services and was working on implementing a mitigation. In a second update at 5:20 PM ET, which the record shows was posted one hour after the first, the company stated: “We have applied the mitigation and are monitoring the recovery.” The incident’s status was listed as Monitoring at that point.

OpenAI’s main status page displayed a “We’re currently experiencing issues” banner alongside the incident and showed it as ongoing for one hour. The company’s status history page lists the same incident under September 17, 2026, with the 5:20 PM update stating that the mitigation had been applied and the recovery was being monitored.

Twelve API Components Listed as Affected

The incident record names 12 affected components, all within the APIs group of OpenAI’s status page: Chat Completions, Responses, Fine-tuning, Embeddings, Images, Batch, Audio, Moderations, Realtime, Files, Login, and Sora. The APIs group comprises 12 components in total on the status page, which tracks the company’s services in separate groups.

Beyond the APIs group, the page lists ChatGPT with 15 components, Codex with four components, FedRAMP with one component, and an Ads Platform group with two components, Ads Manager and the Ads API. The September 17 incident record lists only the API components as affected; no ChatGPT, Codex, FedRAMP, or Ads Platform component appears on the incident’s affected-components list.

Availability Figures Reported on the Status Page

OpenAI’s status page reports uptime figures covering the June 2026–September 2026 window. The page reports the APIs group at 99.93% uptime across its 12 components. Per component, it reports Chat Completions, Responses, Fine-tuning, Embeddings, Batch, Audio, Moderations, Realtime, and Files each at 100% uptime, with Images at 99.95%, Sora at 99.97%, and Login at 99.995%.

For the other groups, the page reports the ChatGPT group at 99.62% uptime across its 15 components, Codex at 100% across its four components, and FedRAMP at 100% for its single component. Within the Ads Platform group, the Ads API is reported at 100% uptime.

OpenAI states on the page that its availability metrics are reported at an aggregate level across all tiers, models, and error types, and that individual customer availability may vary depending on subscription tier and on the specific model and API features in use.

Prior Entries on the Status History

The history page documents a separate incident earlier on September 17, 2026, titled “Elevated errors affecting ChatGPT Work mode,” updated at 4:24 PM ET with the notice that all impacted services had fully recovered. On September 16, 2026, the page records “Elevated errors in ChatGPT Work,” updated at 3:34 PM and likewise marked fully recovered.

On September 15, 2026, the history lists “Elevated errors with gpt-image-2.5-flare,” updated at 8:30 PM, and “Elevated errors from GPT-5.6 and GPT-5.6 Instant on paid plans,” updated at 3:53 AM, each marked fully recovered. Two entries appear on September 14, 2026: “Elevated errors affecting Work Mode in ChatGPT,” updated at 11:11 PM and marked fully recovered, and “Degraded Performance affecting Agents API,” updated at 7:39 PM with the statement that the issue had been resolved and that managed sessions were processing turns normally. A September 13, 2026 entry records “Elevated error rates for Codex and ChatGPT Work,” updated at 11:57 PM and marked fully recovered.

Other API-related entries documented in September 2026 include “Elevated errors for GPT-5.6 Sol on the API” on September 11, updated at 7:21 AM and marked fully recovered, and “Elevated latency in the Responses API” on September 1, updated at 3:05 PM with all impacted services listed as fully recovered.