Is the OpenAI API down? October 21 gpt-5.6-luna errors resolved after 70 minutes
OpenAI reported elevated error rates for gpt-5.6-luna requests on its Chat Completions and Responses APIs on October 21, 2026. The status page timeline runs from 15:34 UTC to full recovery at 16:44 UTC, about 70 minutes of degraded performance.
SessionWatcher editorial · Published · Updated · Source announcement: 2026-10-21
What happened
OpenAI's status page reported elevated error rates for gpt-5.6-luna API requests on October 21, 2026, and classed the incident as degraded performance. It lists two affected components, both under APIs: Chat Completions and Responses.
The affected-component timeline runs from 15:34 UTC to 16:44 UTC, about 70 minutes. At 16:44 UTC OpenAI marked the incident resolved and said all impacted services had fully recovered.
Timeline in UTC
15:34 UTC, Identified: OpenAI said users were seeing elevated errors on the impacted services and that it was working on a mitigation.
15:47 UTC, Monitoring: OpenAI said it had applied the mitigation and was monitoring the recovery.
16:44 UTC, Resolved: OpenAI said all impacted services had fully recovered.
Who was affected
The status page names only the gpt-5.6-luna model and the Chat Completions and Responses APIs. It does not list ChatGPT or Codex as affected components. It also does not mention any change to usage, limits or billing because of the incident.
OpenAI notes that its availability metrics are aggregated across all tiers, models and error types. It says an individual customer's availability may vary with their subscription tier and the model and API features they use.
AI assisted reporting, checked against the linked official sources. Source pages checked 2026-10-21. Editorial process and corrections.