ChatGPT is answering at its usual speed from all 10 countries we time it from.

Our probes' last round was at 14:07 UTC. OpenAI's status page lists ChatGPT as degraded (read 14:16 UTC). Open the status page

ChatGPT over the last 24 hours

Status pageChatGPT

Degraded

Our timingchatgpt.com

On time from 10 countries

Status pageOpenAI API

Operational

Our timingapi.openai.com

On time from 10 countries

The bands in words

OpenAI's status page for ChatGPT: Degraded 13 Nov, 16:11 to 14 Nov, 14:17 UTC. Operational the rest of the time.
Our timing of chatgpt.com: Delayed 13 Nov, 15:37 to 16:22 UTC. On time the rest of the time.
OpenAI's status page for OpenAI API: Operational for the whole 24 hours.
Our timing of api.openai.com: On time for the whole 24 hours.

What the status page listed

The report we read from OpenAI's status page for ChatGPT does not list incidents, so none can be shown here.

The report we read from OpenAI's status page for OpenAI API does not list incidents, so none can be shown here.

What our probes confirmed

chatgpt.com from Mumbai, India (Oracle network)
13 Nov, 15:37 to 16:07 UTC; up to 463 ms against a usual 192 ms, in 2 rounds.

Operational, or on time
Degraded, partial outage or maintenance; or delayed
Major outage, or no answer
No state given, or too few samples
No report read, or no timings
How we measure

chatgpt.com

The front page of chatgpt.com, answered by OpenAI's servers and not from a cache. We do not sign in or send a message, so this is the way in to ChatGPT, not a conversation.

Milliseconds to connect and receive the first byte; looking up the name is not counted. Round of 14:07 UTC.
Timed from	Usual	Now	Reading
Your network
United Statesyour comparison	190	154	On time
Canada	190	168	On time
Brazil	124	152	On time
United Kingdom	172	157	On time
Germany	190	180	On time
France	192	180	On time
India	185	217	On time
Singapore	166	256	On time
Japan	173	158	On time
Australia	140	226	On timelast round slow, not confirmed

api.openai.com

OpenAI's API takes requests at this address. Our request carries no key, so OpenAI's server refuses it at once. The time covers the network path and that refusal. No model runs.

OpenAI's status page lists the OpenAI API as operational (read 14:16 UTC).

Milliseconds to connect and receive the first byte; looking up the name is not counted. Round of 14:07 UTC.
Timed from	Usual	Now	Reading
Your network
United Statesyour comparison	108	86	On time
Canada	118	130	On time
Brazil	196	292	On time
United Kingdom	174	154	On time
Germany	194	192	On time
France	199	150	On time
India	270	270	On time
Singapore	249	269	On time
Japan	166	161	On time
Australia	207	231	On time

Why is ChatGPT slow?

Right now the way in to ChatGPT answers at its usual speed from every country we time it from. If ChatGPT is slow for you, the time is going somewhere our probes do not reach. Slowness can sit in three places.

Your own connection. Wi-Fi, mobile data, a VPN or a company proxy sit between you and ChatGPT, and our probes do not share that path. The button above times the same hosts from your browser: if your times are slow while the probes in your country are not, the delay is on your side.
The answer being written. Our times end at the first byte of the server's reply. A model writes after that, piece by piece, so long answers, hard prompts and modes that reason before answering take longer, and a long conversation gives it more to read before it starts. We time none of this.
The provider. A busy or failing model is for OpenAI to report. Its status page is quoted at the top of this page and part by part below.

OpenAI answers this question for ChatGPT in its help center: Why is my ChatGPT taking so long to respond?

Reading the board

The probes are machines in datacenters, one or two per country. Every 15 minutes they time the same request, and the board shows the middle value. "Usual" is the middle value over the last seven days from the same probes.

The trace draws each round as a multiple of the usual time. The lower line is usual, the upper line is twice usual, and a yellow dot marks a round at or above it. One slow round is common and is not a delay. A country reads "Delayed" only when a probe is that slow in two of three rounds while it times another provider's hosts normally.

OpenAI's status page, part by part

Operational Conversations
Operational Login
Operational ChatGPT Work
Operational Codex in ChatGPT Desktop
Degraded Compliance API
Operational Search
Operational File uploads
Operational Voice mode
Operational GPTs
Operational Image Generation
Operational Deep Research
Operational Agent

This is OpenAI's own statement, read from its public status page. Our timings are not part of it and do not change it.

Confirmed delays for ChatGPT in the last seven days

chatgpt.com from Mumbai, India (Oracle network)
13 Nov, 15:37 to 16:07 UTC; up to 463 ms against a usual 192 ms, in 2 rounds.
chatgpt.com from Bengaluru, India (RACKVOLT CLOUD SOLUTIONS network)
13 Nov, 11:52 to 12:37 UTC; up to 456 ms against a usual 178 ms, in 3 rounds.
chatgpt.com from Bengaluru, India (RACKVOLT CLOUD SOLUTIONS network)
13 Nov, 08:37 to 08:52 UTC; up to 643 ms against a usual 178 ms, in 2 rounds.
api.openai.com from Toronto, Canada (HostPapa network)
12 Nov, 02:07 to 02:22 UTC; up to 306 ms against a usual 120 ms, in 2 rounds.
api.openai.com from Saint-Laurent, Canada (OVH network)
10 Nov, 23:52 to 11 Nov, 00:07 UTC; up to 313 ms against a usual 119 ms, in 2 rounds.
api.openai.com from Mumbai, India (Oracle network)
10 Nov, 21:07 to 21:37 UTC; up to 596 ms against a usual 269 ms, in 2 rounds.
chatgpt.com from Singapore, Singapore (Oracle network)
10 Nov, 17:52 to 18:37 UTC; up to 371 ms against a usual 164 ms, in 2 rounds.
chatgpt.com from Tokyo, Japan (Oracle network)
10 Nov, 13:22 to 13:52 UTC; up to 659 ms against a usual 172 ms, in 2 rounds.