AI API outages on September 25, 2026
UTC day · measured every 5 minutes from 4 regions · method
On September 25, 2026 (UTC) our probes attributed 5 outage(s) to 2 provider(s): Gemini API (Google): 3; Cerebras API: 2. Providers’ own status pages opened 2 entries (Cerebras, OpenAI).
This page is history. Is it down right now?
Live status, updated every 30 minutes: Gemini API (Google), Cerebras API, OpenAI API, iflytek API · all AI APIs, last 72 hours
Was the Gemini API (Google) down on September 25, 2026?
Yes, measured. 3 outage(s) attributed to Google: requests to gemini-flash-lite-latest failed from up to 2 of 4 regions at once while no other provider was failing, first at 11:20 UTC, at least 2 minutes observed in total.
| UTC | What failed | Regions | Duration | Failed probes | Verdict | Error |
|---|---|---|---|---|---|---|
| 11:20–11:20 | requests to gemini-flash-lite-latest | 2 of 4: South America (São Paulo), US (Central) | <5 min | 2 | provider outage | timeout |
| 14:18–14:20 | requests to gemini-flash-lite-latest | 2 of 4: Europe (Germany), US (Central) | ≥2 min | 2 | provider outage | timeout |
| 17:40–17:40 | requests to gemini-flash-lite-latest | 2 of 4: South America (São Paulo), US (Central) | <5 min | 2 | provider outage | timeout |
| 17:56–18:00 | requests to gemini-flash-lite-latest | 4 of 4: Asia (Tokyo), Europe (Germany), South America (São Paulo), US (Central) | ≥4 min | 4 | shared with 1 other provider(s) — not attributed | http 503; timeout |
| 18:16–18:20 | requests to gemini-flash-lite-latest | 3 of 4: Asia (Tokyo), Europe (Germany), South America (São Paulo) | ≥4 min | 3 | shared with 1 other provider(s) — not attributed | timeout |
Current Gemini API (Google) status
Was the Cerebras API down on September 25, 2026?
Yes, measured. 2 outage(s) attributed to Cerebras: host probes failed from up to 4 of 4 regions at once while no other provider was failing, first at 12:29 UTC, at least 6 minutes observed in total. Cerebras’s own status page opened 1 entry that day (self-reported, listed below).
| UTC | What failed | Regions | Duration | Failed probes | Verdict | Error |
|---|---|---|---|---|---|---|
| 12:29–12:35 | host probes | 4 of 4: Asia (Tokyo), Europe (Germany), South America (São Paulo), US (Central) | ≥6 min | 4 | provider outage | timeout |
| 12:53–12:55 | host probes | 2 of 4: Europe (Germany), US (Central) | ≥2 min | 2 | shared with 1 other provider(s) — not attributed | timeout |
| 13:20–13:20 | host probes | 2 of 4: Asia (Tokyo), US (Central) | <5 min | 2 | provider outage | timeout |
Self-reported on Cerebras’s status page, not verified by our probes:
- major Partial Service Disruption · resolved Sep 25, 17:14 UTC
Was the OpenAI API down on September 25, 2026?
Not by our measurements. No failure of OpenAI was seen from two or more regions at once. OpenAI’s own status page opened 1 entry that day (self-reported, listed below).
Self-reported on OpenAI’s status page, not verified by our probes:
- critical Issues with Codex · resolved Sep 25, 23:54 UTC
Was the iflytek API down on September 25, 2026?
Not attributably. 1 multi-region failure(s) of iflytek coincided with failures of other providers, so the cause sat above iflytek and is not blamed on it.
| UTC | What failed | Regions | Duration | Failed probes | Verdict | Error |
|---|---|---|---|---|---|---|
| 15:25–15:35 | host probes | 2 of 4: Asia (Tokyo), South America (São Paulo) | ≥10 min | 2 | shared with 2 other provider(s) — not attributed | timeout |
63 single-region failure(s) were also recorded that day (google 10, baichuan 9, aleph-alpha 6, upstage 6, nebius 5, deepinfra 5, iflytek 4, deepseek 4, doubao 3, sensenova 3). A failure seen from one region cannot be separated from a problem on that network path, so none of them is called an outage.
How is an outage told apart from a network problem?
Each provider is probed every five minutes from 4 independent regions. A failure is attributed to the provider only when it is seen from two or more regions at once and no other provider fails in the same minutes; a failure shared with other providers has a cause above all of them and is not blamed on anyone. Durations are lower bounds: we see the first and last failed probe, not the real start and end. Status-page entries are the provider’s own words, shown next to the measurement, not instead of it.
Previous: Sep 24 · Last 72 hours · Next: Sep 26
Machine-readable: incidents.json · Licence: CC BY 4.0