How to Test Proxy Speed: Latency, Throughput and Success Rate

A proxy that pings fast can still be slow on your real target. Here is how to measure the numbers that matter, with ready-to-use commands and a simple way to compare providers fairly.

By Tomas LindqvistUpdated: 5 min read
How to test proxy speed — measuring latency, time to first byte and success rate through a proxy

To test proxy speed properly, measure four things on the website you actually care about: connection latency, time to first byte (TTFB), throughput for larger downloads and the success rate across many requests. A single ping or a generic speed-test page tells you very little, because the route to your real target, the target's own response time and how it treats the proxy's IP all change the result.

This guide explains each metric, gives you copy-paste commands for curl and a small script, and shows how to compare proxies or providers fairly.

The four metrics that matter

MetricWhat it measuresWhy it matters
Connect timeTime to open a connection through the proxyShows distance and load between you, the proxy and the target
TTFBTime until the first byte of the response arrivesClosest to how fast pages feel; includes the target's processing time
ThroughputDownload speed on a larger fileMatters for media, large pages and bulk transfers
Success rateShare of requests that return the expected responseA fast proxy that gets blocked half the time is a slow proxy
Key takeaway: success rate is usually more important than raw speed. Always test on your real target, not just on a neutral test page.

Quick test with curl

curl is installed on most systems and can print detailed timings. Replace the placeholders with your proxy credentials and target URL:

curl -x http://USER:PASS@HOST:PORT -o /dev/null -s \
  -w "connect: %{time_connect}s  ttfb: %{time_starttransfer}s  total: %{time_total}s  code: %{http_code}\n" \
  https://example.com/

For a SOCKS5 proxy, use -x socks5h://USER:PASS@HOST:PORT. The h makes the proxy resolve the domain name, which is usually what you want.

Read the output like this: connect should be low and stable, ttfb is the number to compare between proxies, and code should be 200 (or whatever your page normally returns). A 403, 429 or a CAPTCHA page means the request was technically fast but practically failed.

Testing many requests

One request is noise. Run at least 20–50 requests per proxy to get a reliable picture. A simple loop in a terminal:

for i in $(seq 1 30); do
  curl -x http://USER:PASS@HOST:PORT -o /dev/null -s \
    -w "%{http_code} %{time_starttransfer}\n" https://example.com/
done

From the results, calculate:

  • Success rate — the share of lines with the expected status code.
  • Median TTFB — more meaningful than the average, which a few slow requests can distort.
  • 95th percentile — the time that 95% of requests beat. It shows how bad the slow cases are.

Measuring throughput

If you download large pages, images or files, also test sustained speed. Point curl at a file of a few megabytes on a server you are allowed to use and read %{speed_download} (bytes per second). Keep in mind that residential and mobile proxies are billed per GB, so heavy throughput tests consume paid traffic; a few runs are enough.

Reading common errors in your results

Failed requests tell you as much as fast ones. When you go through the output, group failures by type:

  • 407 Proxy Authentication Required — wrong username or password, or your IP is not on the allowed list. Fix the credentials before testing further.
  • 403 or a CAPTCHA page — the target recognised automated traffic or distrusts the IP range. Try a different proxy type or slow down.
  • 429 Too Many Requests — you hit the site's rate limit. Reduce concurrency and add delays.
  • Timeouts — the proxy or the target did not respond in time. A few are normal for residential pools; many point to a routing or capacity problem.
  • Connection refused — wrong host or port, or the protocol does not match (for example HTTP settings on a SOCKS5 port).

How to compare proxies fairly

  1. Use the same target — test every proxy against the same URLs you will use in production.
  2. Test at the same time — network conditions change during the day; run comparisons side by side.
  3. Match locations — a proxy close to the target server will usually beat one on the other side of the world.
  4. Use enough requests — at least 30 per proxy, more for rotating pools.
  5. Compare medians and success rate — not the single best result.
  6. Repeat later — run the test again on another day before deciding.
Tip: test from the machine where your tool will actually run. A proxy that looks slow from your laptop on home Wi-Fi may be fast from a server in a data centre, and vice versa.

What affects proxy speed

  • Proxy type — datacenter and ISP proxies are usually fastest; residential and mobile depend on the end user's connection.
  • Distance — the path from you to the proxy and from the proxy to the target both add latency.
  • Rotation — a new IP on each request means a new connection each time; sticky sessions reuse connections and can be faster. See rotating vs sticky proxies.
  • Target behaviour — rate limits, bot protection and slow back ends affect every proxy.
  • Your own code — reusing connections, limiting concurrency sensibly and not downloading unnecessary assets often matter more than the proxy.

Typical results to expect

There are no universal numbers, but as a rule of thumb datacenter and ISP proxies in the same region as the target often return the first byte noticeably faster than residential or mobile proxies, which route through consumer connections. Residential proxies trade some speed for a much higher success rate on strict sites. If your success rate is high and TTFB is stable, the proxy is doing its job, even if a speed-test site shows a modest number. For scraping-specific tuning, read our guide to proxies for web scraping.

Watch out: load-testing a website you do not own, or sending large bursts of requests to measure speed, can harm the site and breach its terms. Keep tests small, respect rate limits and only benchmark against targets you are allowed to use.

Test mistPROXY on your own targets

mistPROXY offers residential proxies in 190+ countries, plus ISP, datacenter IPv4, IPv6 and mobile proxies, all with SOCKS5 and HTTP(S). Credentials appear instantly after purchase, so you can start a small test within minutes and scale only after the numbers convince you.

99.9%network uptime
190+countries
Instantcredentials

See how it works, pick a plan on the pricing table or create a free account. Questions about replacements are answered in the FAQ and on Telegram at @mistnetwork.

Frequently asked questions

What is a good proxy response time?
It depends on type and distance. Compare median time to first byte on your real target; a stable TTFB with a high success rate matters more than a single fast result.
How do I test a proxy with curl?
Run curl with -x pointing to the proxy and -w printing time_connect, time_starttransfer and http_code. Repeat 20–50 times and compare medians.
Why is my proxy fast on a speed test but slow on my site?
The target may be far from the proxy, slow to respond or rate-limiting the IP. Always measure on the site you actually use.
Are datacenter proxies faster than residential?
Usually yes in raw speed, but residential proxies often succeed more on strict sites. Success rate often decides which is faster in practice.
Does speed testing use residential traffic?
Yes. Residential and mobile proxies are billed per GB, so keep tests small and avoid large file downloads.

Related articles

Latest articles

All articles