What Really Happened During OpenAI's May 2026 Service Slowdown

User feedback described response times as noticeably slower from the early morning hours of May 27 Beijing time. While the core services remained partially functional, the latency was significant enough to disrupt normal workflows. Third-party monitor 9to5Mac provided a running timeline on May 27, noting that ChatGPT’s latency issue was resolved by mid-morning Pacific Time, while the API slowdown persisted for several more hours .

Which components were affected?

Unlike a previous April 2026 partial outage that knocked out 12 ChatGPT components and one Codex component , the May 27–28 incident appeared more targeted. Independent monitoring platform Pagerly showed one or more OpenAI components experiencing issues . Additional third-party services like StatusGator recorded 22 user-submitted outage reports in a 24-hour period, with their dashboard confirming that the OpenAI APIs issue was eventually resolved .

Even after the main latency incident was resolved, OpenAI’s status page continued to display lingering problems—specifically, a Codex context-compression issue running slower than expected and a problem with Android ChatGPT Enterprise when switching workspaces . Whether these were directly related to the same root cause remains unclear from the public record.

Confirmed timeline of the incident

While the exact minute-by-minute resolution is not documented in exhaustive detail across all independent monitors, the following sequence is verifiable from the source record:

• May 27, 2026 (Beijing morning): Users begin reporting elevated latency on ChatGPT and APIs .
• May 27, 2026, 22:47 (Beijing): OpenAI officially acknowledges the “high latency” status on its page .
• May 28, 2026, 04:06 (Beijing): OpenAI marks the latency issue as resolved, about 5 hours and 19 minutes after the official acknowledgment .
• Post-resolution: Codex context-compression sluggishness and Android Enterprise workspace switching issues remain listed on the status page .
• May 28, 2026, ongoing: StatusGator logs warning periods of 23 hours 59 minutes, 1 day, and 1 hour 34 minutes across consecutive days for OpenAI APIs .

The GPT-5.5 Instant update: related or coincidence?

On the same day the latency disruption was being resolved—May 28, 2026—OpenAI’s public API changelog announced a new chat-latest snapshot that points to the latest Instant model currently used in ChatGPT, recommending GPT-5.5 for production API usage .

It is important to note that this was not the initial launch of GPT-5.5 Instant. That occurred weeks earlier on May 5, 2026, when GPT-5.5 Instant replaced GPT-5.3 Instant as ChatGPT’s default model . The May 28 changelog entry therefore represents a routine snapshot update rather than a novel model deployment.

No causal connection has been established in any of the provided sources linking the chat-latest change to the latency incident. The proximity of the two events has prompted speculation among users, but the public record treats them as separate activities with overlapping timing.

How this fits into OpenAI’s 2026 outage pattern

The May latency degradation did not happen in isolation—2026 has been a notably bumpy year for OpenAI’s service reliability. A look at previous incidents shows a pattern of escalating and varied disruptions.

February 2026 cluster

According to outage-history tracker apistatuscheck.com, 21 incidents were recorded for OpenAI in February 2026 alone . Other monitoring services like Liputan6 and 9to5Mac described major ChatGPT outages on February 3, affecting logged-out users, login functionality, content loading, and “Try again” buttons . Downdetector recorded a surge from dozens to over 13,000 user reports during that event .

April 2026 partial outage

On April 20, 2026, OpenAI experienced a partial outage that affected 12 ChatGPT components and one Codex component—including login, voice mode, and search functionality . Downdetector recorded more than 7,600 reports in the UK and 1,700 in the US during that disruption .

March 2026 Codex degraded performance

OpenAI’s official status page documents a separate Codex degraded-performance incident from March 11, 2026, where mitigation took several hours and required renewed investigation after user reports resurfaced . This incident is distinct from the May latency event, but it reinforces that Codex reliability has been a recurring theme in 2026.

May 2026 context

Compared to the February and April blackouts, the May 27–28 event was more surgical: mainly high latency rather than outright inaccessibility. Nonetheless, the accumulation of incidents—each varying in scope and resolved without a public root-cause analysis—points to systemic growing pains as OpenAI scales its infrastructure.

Conclusion: what we still don’t know

As of the most recent available information in late May 2026, OpenAI has not publicly attributed a specific root cause for the May 27–28 elevated latency. The company acknowledged and resolved the issue within hours, and the lingering Codex compression slowness suggests that some backend challenges persisted. The coincidental timing of the chat-latest update remains just that—coincidental—until any official technical correlation is provided. For now, the incident serves as another data point in a year where reliability has become a closely watched metric for the AI industry’s most visible platform.