Over the 90 days to November 13, 2026, Anthropic's status page showed 99.54% uptime for claude.ai and OpenAI's showed 99.26% for the ChatGPT group. Google and Microsoft do not publish uptime percentages for Gemini or Copilot. The week of November 9 was the busiest of the autumn: a Microsoft 365 outage took Copilot down for most of a day, and Claude had two major incidents the following morning.
Published Uptime, Past 90 Days
| Assistant | Component | Uptime |
|---|---|---|
| Claude | claude.ai | 99.54% |
| Claude | Claude API | 99.77% |
| Claude | Claude Code | 99.69% |
| ChatGPT | ChatGPT group, all components | 99.26% |
| ChatGPT | Conversations | 99.95% |
| ChatGPT | Login | 99.93% |
| OpenAI | API group | 99.97% |
| Gemini | Not published | Not published |
| Microsoft Copilot | Not published | Not published |
The two companies calculate these differently. OpenAI's ChatGPT group figure is lower than any of the main components listed under it, so the group number and the component numbers are not the same measure. Treat cross-company comparisons as approximate.
The Week of November 9, 2026
| Date (UTC) | Assistant | What happened |
|---|---|---|
| Nov 9, 23:08 to 23:50 | OpenAI API | Elevated errors on 14 components. No cause published |
| Nov 11, 20:43 to 20:52 | ChatGPT | Errors on logins, sign-ups and several features |
| Nov 12 to 13 | Microsoft Copilot | Microsoft 365 Copilot was part of the Microsoft 365 outage. Users could be unable to load Copilot Chat or reach email, calendar and files |
| Nov 12, from 17:52 | Claude | Microsoft 365 connector errors. Anthropic said the cause was Microsoft's incident. Core Claude was not affected that day |
| Nov 12, from 18:29 | ChatGPT | Microsoft connector errors, rated minor. No conversation or login incident listed |
| Nov 13, 09:58 to 11:58 | Claude | Major incident. Authentication problems on claude.ai, Claude Code and the console |
| Nov 13, 13:09 to 15:38 | Claude | Major incident. Elevated errors on claude.ai, the API and Claude Code, spreading to connectors and mobile apps |
Anthropic tied only the connector incident to Microsoft. It published no cause for the two November 13 incidents.
Incidents Posted Per Month, 2026
| Month | Claude, all posts | Claude, major or critical | OpenAI, with component impact |
|---|---|---|---|
| January | 26 | 6 | 14 |
| February | 58 | 15 | 21 |
| March | 51 | 21 | 33 |
| April | 49 | 16 | 25 |
| May | 41 | 17 | 25 |
| June | 47 | 9 | 29 |
| July | 55 | 17 | 31 |
| August | 25 | 8 | 18 |
| September | 13 | 8 | 30 |
| October | 17 | 11 | 37 |
| November, to the 13th | 5 | 3 | 12 |
| Total | 387 | 131 | 275 |
An incident is a post on a status page, not a measure of downtime. Anthropic posts many small, model-specific notices, so a higher count does not mean more time down. Of OpenAI's 275 incidents, 6 reached a full outage of a component and 32 a partial outage.
Notable Outages by Assistant
| Assistant | Date | Length | Published cause |
|---|---|---|---|
| Claude | Mar 25 to 27 | Three periods over three days | Network degradation inside its serving infrastructure |
| Claude | Apr 15 | About 2 hours 50 minutes | Not published |
| ChatGPT | Jul 23 | About 6 hours 40 minutes | A network maintenance failure at an infrastructure provider |
| ChatGPT | Apr 20 | About 1 hour 40 minutes | An overloaded internal authentication dependency |
| Gemini | Jun 10 | About 6 hours 55 minutes of elevated errors | Database read contention. Google reported a 50% error rate on prompts |
| Gemini | Feb 18 to 22 | About 4 days | An internal job deleted conversation metadata, so chat history was not visible. Restored from backups |
| Copilot | Nov 12 to 13 | More than 20 hours | Part of the Microsoft 365 outage. Root cause not published |
Full lists: Claude outage history and ChatGPT outage history.
What Causes Them
- Configuration changes lead. Most of OpenAI's 2026 write-ups name a configuration change or rollout.
- Login is the weak point. Authentication appears in outages at Anthropic, OpenAI and Microsoft this year. When login fails, the model can be healthy and still unreachable.
- Dependencies spread failures. On November 12, a Microsoft outage broke features inside Claude and ChatGPT. OpenAI traced several July incidents to infrastructure providers.
- Transparency varies. OpenAI published 24 incident write-ups in 2026. Anthropic published eight, none after March. Google published full reports for two Gemini incidents. Copilot has no public incident history.
If Your Team Depends on One Assistant
- Keep a second assistant ready. All four had multi-hour incidents in 2026.
- Watch the official status page. Crowd-sourced report counts rise with news coverage and are not a measure of impact.
- Separate chat from API. Consumer apps and APIs often fail independently. See LLM API uptime and reliability.
Methodology
Claude data comes from status.claude.com and its incident history. OpenAI data comes from status.openai.com. Gemini incidents come from the Google Workspace Status Dashboard, whose feed ended at September 25, 2026 when read. Copilot has no readable official history, so its entries come from press coverage of Microsoft's advisory. All status pages were read on November 13, 2026 at about 17:20 UTC. Incident counts are our tally of status page posts. Presenc AI did not measure uptime independently.