Benchmark · measured, not promisedevery 15 min
99.70% success rate over 30 days.
Every 15 minutes we request a fixed list of popular sites through the same gateway customers use, as an ordinary customer account, and publish every result for the listed targets — the failures included.
99.70%success rate · 30 days
100.00%connect rate · 30 days
494 msmedian time to first byte
20,160requests measured
20,100 of 20,160 requests succeeded; 60 failed and are counted. 95% confidence interval: 99.61% – 99.77%. Last measured 9 Nov 2026, 05:33 UTC · measuring since 26 Sept 2026, 10:55 UTC.
By time window.
All targets together| Window | Success rate | Connect rate | TTFB median | TTFB p95 | Requests | Failed |
|---|---|---|---|---|---|---|
| Last 24 hours | 98.80% | 100.00% | 498 ms | 820 ms | 672 | 8 |
| Last 7 days | 99.68% | 100.00% | 493 ms | 911 ms | 4,704 | 15 |
| Last 30 days | 99.70% | 100.00% | 494 ms | 894 ms | 20,160 | 60 |
By site.
One GET per site per run · the exact URL we request| Site | Success · 30 days | Success · 24 hours | TTFB median | Requests · 30 days |
|---|---|---|---|---|
| Amazon.comwww.amazon.com · No measurements yet | — | — | — | 0 |
| bbcwww.bbc.com/news | 99.96% | 100.00% | 391 ms | 2,880 |
| bingwww.bing.com | 99.96% | 100.00% | 501 ms | 2,880 |
| githubgithub.com | 99.72% | 95.83% | 490 ms | 2,880 |
| googlewww.google.com | 99.02% | 100.00% | 403 ms | 2,880 |
| walmartwww.walmart.com | 99.96% | 100.00% | 599 ms | 2,880 |
| wikipediaen.wikipedia.org/wiki/Main_Page | 99.96% | 100.00% | 454 ms | 2,880 |
| zillowwww.zillow.com | 99.30% | 95.83% | 611 ms | 2,880 |
Every failure, by cause.
Last 30 days · counted in every rate above- 47Blocked or challengedSite side. 401, 403 or 429, a captcha or challenge page, or a redirect to one.
- 9Site timed outSite side. The tunnel was up, but the site did not answer in time.
- 4Other HTTP errorSite side. Any other status outside 2xx and 3xx, such as 404 or 500.
How we measure.
So you can check the number, or run it yourselfWhat we send
- Every 15 minutes, one GET request to each site in the table above — the front page or a public page, never anything behind a login.
- Through the same production gateway customers use, as an ordinary customer account — no private route, no hand-picked exit.
- A new connection for every request, a desktop browser User-Agent, a 30-second timeout, redirects not followed.
- No retries. Each attempt is recorded once, the moment it finishes.
What counts as a success
- The site answered with HTTP 2xx or 3xx,
- and the page is not a block, captcha or challenge page (Cloudflare, AWS WAF, DataDome, PerimeterX, Akamai and the sites' own robot checks) — including the ones served with a 200 or 202 status,
- and a redirect does not point at one.
What counts as a failure
- Everything else: 401, 403 and 429, any other 4xx or 5xx, a challenge page served as 200, timeouts, dropped connections, and every case where the gateway could not connect at all.
- For the published targets, failures are never removed. They are in every rate on this page and listed by cause above.
The other numbers
- Connect rate: the share of requests where the gateway opened the tunnel to the site, whatever the site then answered.
- Time to first byte: from sending the request to the first byte of the answer, including connecting through the gateway. Successful requests only; median and 95th percentile.
- Rates are rounded down to two decimals: 99.996% is shown as 99.99%, never as 100.00%.
What this does not measure
- Amazon.com is included in the published target list. Our plain HTTP client may receive an AWS WAF JavaScript challenge or a robot check; these count as failures, including HTTP 200 or 202 challenge pages. We do not solve CAPTCHAs. Booking and IMDb remain outside this list. Retired targets are excluded from all public rates and failure counts; their stored history is not reassigned to replacement sites.
- Your own targets, volumes and request patterns. A front page every 15 minutes is gentler than a crawl; test your workload on the free traffic before you rely on a number.
- A run cut short by our own measurement job restarting is discarded rather than recorded as failures. An outage of the gateway itself is recorded, like any other failure.
Measure it on your own target.
HTTP or SOCKS5 on one endpoint. 1 GB free · no card.
Get 1 GB free