Skip to content

Live benchmarks

Our real numbers,
including the bad ones.

Every proxy vendor advertises 99% uptime, which measures whether their servers are switched on. This page measures something you can actually spend: whether a request to a specific target came back with the page you asked for. Updated hourly, method published, nothing excluded.

Last updated 1,164,000 requests in this window

46.0M

Total live pool

46,022,130 IPs

41.3M

Residential

89.7% of pool

3.9M

Mobile

8.6% of pool

612K

ISP

1.3% of pool

185K

Datacenter

0.4% of pool

Per-target success rate

Last completed hour, by target and pool

Sorted best to worst. We are not going to bury the bottom of this list — if a target is hard right now, knowing that before you buy is worth more to you than a flattering average.

  • EtsyResidential
    98.4%
  • AmazonResidential
    98.1%
  • Booking.comResidential
    97.9%
  • Google SearchResidential
    97.4%
  • TargetResidential
    97.2%
  • WalmartResidential
    96.8%
  • YouTubeResidential
    96.1%
  • ZillowISP
    95.6%
  • InstagramMobile
    94.2%
  • IdealistaResidential
    93.8%
  • TikTokMobile
    92.7%
  • LinkedInISP
    89.3%

Network average 96.4%Axis starts at 80% — differences look larger than they are.

  • Strong≥ 96%
  • Steady92–96%
  • Variable< 92%

Weighted average

96.4%

Across every target on this page

Median TTFB

640ms

Time to first byte, all pools

Hardest target

89.3%

LinkedIn on ISP

Thirty days

It moves, and we show it moving

A flat line on a vendor's chart usually means the chart is decorative. Ours dips when a major target changes its defences, and recovers when we adapt.

94%95%96%97%98%2026-07-082026-08-06
Network-wide weighted success rate, 30 days. Axis clipped to 94–98%.

Full record

Every measurement in this window

Sample size included, because a 99% success rate over 40 requests is not a measurement — it is a coincidence.

Per-target proxy success rate, median time to first byte, sample size and 24-hour change
TargetPoolCategorySuccessMedian TTFBRequests24h
Etsyetsy.comResidentialRetail98.4%480ms67,200+0.2
Amazonamazon.comResidentialRetail98.1%520ms184,200+0.3
Booking.combooking.comResidentialTravel97.9%560ms88,100+0.4
Google Searchgoogle.comResidentialSearch97.4%410ms221,800−0.2
Targettarget.comResidentialRetail97.2%590ms71,400+0.6
Walmartwalmart.comResidentialRetail96.8%610ms112,300+0.1
YouTubeyoutube.comResidentialMedia96.1%690ms129,500−0.5
Zillowzillow.comISPListings95.6%700ms53,600+0.8
Instagraminstagram.comMobileSocial94.2%780ms96,400+1.1
Idealistaidealista.comResidentialListings93.8%820ms22,900−0.9
TikToktiktok.comMobileSocial92.7%860ms74,900−1.4
LinkedInlinkedin.comISPSocial89.3%940ms41,700−2.6

Methodology

How these numbers are produced

Published in full so you can argue with it. If you think the method flatters us, tell us where and we will fix it or explain why not.

What counts as a success

An HTTP 2xx response carrying the page body we expected, returned inside 30 seconds. A CAPTCHA page, a block page, an empty 200 and a timeout all count as failures. We do not retry before recording the result.

How often we measure

Every target is sampled continuously from every region we sell. Figures on this page cover the last completed hour and are republished at the top of the next one.

Which pool each number comes from

Each row names the pool it was measured on. Mixing pools would flatter the average, so we do not. Where a target behaves very differently across pools, we list it once per pool.

What we do not do

We do not hand-pick regions, drop the worst hour, exclude failed requests from the denominator, or warm up sessions before measuring. When a number gets worse, it gets published worse.

Use this data. Published under CC BY 4.0. Cite ProxyHive and link back to this page. If you want the underlying time series rather than the summary, email [email protected] and we will send it.

Questions

About these benchmarks

How often are these proxy benchmarks updated?
Every hour. Each figure covers the last completed hour of measurement and is republished at the top of the next one, whether the number improved or got worse.
What counts as a successful request?
An HTTP 2xx response carrying the page body we expected, returned within 30 seconds. CAPTCHA pages, block pages, empty 200s and timeouts are all recorded as failures, and we do not retry before recording the result.
Why do success rates differ between pools?
Because targets treat network origin differently. A site that waves through a mobile carrier address may block a datacenter range outright. Each row names the pool it was measured on so the comparison is like for like.
Can I use this data?
Yes. It is published under CC BY 4.0 — cite ProxyHive and link back. If you want the raw series rather than the page, email us and we will send it.
Do you exclude bad results?
No. We do not hand-pick regions, drop the worst hour, remove failed requests from the denominator, or warm up sessions before measuring. Targets we perform badly on stay on this page.

Test it against your own targets

1 GB free, no card. Point it at whatever you actually scrape and compare your numbers to ours. If they do not match, we want to hear about it.

Figures on this page are placeholders wired to the same shape as the live measurement pipeline. See the product pages for what each pool is built for.