Is Groq Down?

No — Groq is up

Reachable from all 8 checked regions

Average response time: 132ms

Last checked 3 hours ago · checks run every 6 hours

Official status page: https://groqstatus.com

Groq uptime

100%

Last 7 days

100%

Last 30 days

100%

Last 90 days

181ms

Avg response, 30 days

Measured from multiple regions every 6 hours. Percentages count only checks that returned an availability answer — 64 days measured so far. A dash means that window does not yet hold enough measured days to publish a figure.

30-day history

64-day clean streak

Sep 14 Today

No downtime Partial Downtime Not measurable No data

Reachability by region

Each region runs its own request from a different part of the world. A service can be up for one continent and down for another, which is usually the first sign of a routing or CDN problem.

ams

99ms

DNS 23ms TCP 4ms TLS 20ms TTFB 95ms

arn

125ms

DNS 46ms TCP 4ms TLS 35ms TTFB 117ms

cdg

118ms

DNS 36ms TCP 10ms TLS 15ms TTFB 111ms

ord

109ms

DNS 31ms TCP 1ms TLS 24ms TTFB 108ms

sin

139ms

DNS 92ms TCP 1ms TLS 6ms TTFB 137ms

sjc

191ms

DNS 18ms TCP 1ms TLS 9ms TTFB 183ms

syd

73ms

DNS 6ms TCP 1ms TLS 7ms TTFB 72ms

yyz

203ms

DNS 29ms TCP 12ms TLS 33ms TTFB 191ms

What Groq does

Groq runs large language models on its own inference hardware and sells access through GroqCloud, an API aimed at applications that need fast responses. Developers point existing model calls at it for speed, and it serves both production and preview models. An application built around its latency has no slower fallback unless one was deliberately written in.

What an outage looks like

Groq separates the website from the API and reports production and preview models as their own components, so one model can be degraded while the rest of the platform is fine. Requests fail as 5xx errors or time out, and a streamed response can stop partway rather than returning a clean error, which surfaces in an application as a truncated answer instead of a failure.

What to do about it

Check groqstatus.com, which reports the API and individual production and preview models separately, so confirm the specific model you call rather than reading the top-level banner. It publishes uptime per component over a rolling multi-month window. If one model is affected, switching to another on the same API is usually faster than waiting, provided your prompts are not tuned tightly to one model.

Is it down for everyone, or just you?

If this page says Groq is up but it is not loading for you, the problem is between you and them. Run a check against any URL from up to 17 cities across three regions to find out where it breaks.

Test it yourself

Related services

Groq outage FAQ

The API is up but my model returns errors. Is that expected?

Yes, that pattern is normal on Groq's status page, which lists individual production and preview models as separate components under the API. One model can be degraded or withdrawn while the API itself is healthy. Check the line for the exact model string in your request, since the top-level banner covers GroqCloud broadly rather than the model you happen to call.

Why did my response stop halfway through?

Streamed responses can stop partway during a degradation rather than returning an error, so the application sees a short or truncated answer instead of a failure it can retry. If you are seeing unexplained truncation, check the status page before assuming a prompt or token-limit problem. Applications that stream should treat an unexpectedly terminated stream as a failure rather than as a complete answer.

Should I switch models during a Groq incident?

If a single model is affected and others are healthy, switching is usually faster than waiting, since the components are tracked independently. The caveat is that prompts tuned tightly to one model may behave differently on another, so this works best when your prompts are reasonably portable. Preview models are listed separately from production ones and are the more likely to change.

What is the difference between preview and production models?

Groq lists them as separate components on its status page, which is the practical signal: production models are the ones intended for stable use, and preview models are earlier in their lifecycle. If your application depends on a preview model, its status line is the one to watch, and its behaviour is more likely to change over time than that of a production model.

How we measure this

We request Groq's public endpoint every 6 hours from Fly.io regions across six continents — the most recent check ran from 8 of them.
A region counts as down only when it gets no usable HTTP response. A 403 or 429 means the origin answered and refused us, which we report as blocked, never as an outage.
A single failing region is treated as probe noise. We only change the verdict when two consecutive cycles agree.
Response times average only the regions that actually served the page, so a timeout never inflates the number.
Where Groq publishes an official status feed we read it too. An all-clear from the vendor can soften an unconfirmed degradation; a vendor-declared outage only worsens our verdict when our own checks corroborate it.

Get alerted when Groq goes down

This page refreshes every 6 hours. Your own monitors run as often as every 30 seconds, from up to 17 cities across three regions, and tell you the moment something breaks.

Start Free Monitoring

Free plan available No credit card required