logo
  • Proxy selection
    • By type
    • By country
    • By purpose
  • Rating
  • Proxy Providers
  • Blog
  • For Companies
  • Tools
    • Proxy Checker
    • WebRTC Leak

Proxy benchmark: our test results

We buy or get access to proxies and test them ourselves from a server in London: 200 requests per rotating proxy type, 50 per static type, repeated every week while we have access. This page puts every provider we have measured into one table per proxy type. Last run: October 6, 2026. 6 providers, 11 test runs, method version 2026-09-25.5.

This is not a ranking: rows are in alphabetical order, and only providers we have actually tested are listed. Our ranked lists weigh price, coverage and features as well. How each number is calculated: formulas and test code on GitHub.

Rotation Residential Proxies 5Rotation Mobile Proxies 2Shared Datacenter Proxies 2Individual IPv6 Proxies 1Dedicated Datacenter Proxies 1

Rotation Residential Proxies

Among the 5 providers we tested:

  • highest success rate — Evomi, Thordata (100%)
  • fastest response on a new connection — Evomi (802 ms)
  • fastest response on an open connection — NodeMaven (207 ms)
  • fastest download — ThunderProxy (13.1 Mbit/s)
  • fewest datacenter IPs — Evomi, Thordata, ThunderProxy (0%)
Response time of every requestWhere the response times of 200 requests fall. Log scale: further right is slower. The number is the median.
2005001k2k5k10k20kEvomi: median 802 ms, 90% < 1624 ms, 200/200Evomi802NodeMaven: median 1248 ms, 90% < 2860 ms, 199/200NodeMaven1248Proxmint: median 1157 ms, 90% < 2399 ms, 198/200Proxmint1157Thordata: median 2168 ms, 90% < 2839 ms, 200/200Thordata2168ThunderProxy: median 1289 ms, 90% < 2053 ms, 198/200ThunderProxy1289
2005001k2k5k10k20kEvomi: median 802 ms, 90% < 1624 ms, 200/200Evomi802NodeMaven: median 1248 ms, 90% < 2860 ms, 199/200NodeMaven1248Proxmint: median 1157 ms, 90% < 2399 ms, 198/200Proxmint1157Thordata: median 2168 ms, 90% < 2839 ms, 200/200Thordata2168ThunderProxy: median 1289 ms, 90% < 2053 ms, 198/200ThunderProxy1289
middle half of requests80% of requestsmedianms
Key metrics side by side
Response on a new connection ms, lower is better
Evomi802
NodeMaven1248
Proxmint1157
Thordata2168
ThunderProxy1289
Response on an open connection ms, lower is better
Evomi251
NodeMaven207
Proxmint585
Thordata604
ThunderProxy413
Speed Mbit/s, higher is better
Evomi2.5
NodeMaven3.4
Proxmint7.9
Thordata3.7
ThunderProxy13.1
Zillow %, higher is better
Evomi87.5
NodeMaven50
Proxmint87.5
Thordata75
ThunderProxy87.5
ProviderSuccessResponse, msnew / openSpeedMbit/sBlocklistsDatacenter IPsZillowRedditTestedData
Evomi100%802 / 2512.50%0%87.5%100%September 25, 2026CSV
NodeMaven99.5%1248 / 2073.40%1%50%100%September 25, 2026CSV
Proxmint99%1157 / 5857.92.5%1.5%87.5%100%October 6, 2026CSV
Thordata100%2168 / 6043.70%0%75%100%September 25, 2026CSV
ThunderProxy99%1289 / 41313.10%0%87.5%100%October 2, 2026CSV

Rotation Mobile Proxies

Response time of every requestWhere the response times of 200 requests fall. Log scale: further right is slower. The number is the median.
2005001k2k5k10k20kEvomi: median 904 ms, 90% < 2008 ms, 199/200Evomi904NodeMaven: median 1366 ms, 90% < 3124 ms, 194/200NodeMaven1366
2005001k2k5k10k20kEvomi: median 904 ms, 90% < 2008 ms, 199/200Evomi904NodeMaven: median 1366 ms, 90% < 3124 ms, 194/200NodeMaven1366
middle half of requests80% of requestsmedianms
ProviderSuccessResponse, msnew / openSpeedMbit/sBlocklistsDatacenter IPsZillowRedditTestedData
Evomi99.5%904 / 2613.30%0%100%100%September 25, 2026CSV
NodeMaven97%1366 / 5424.30.5%0%87.5%100%September 25, 2026CSV

Shared Datacenter Proxies

Response time of every requestWhere the response times of 200 requests fall. Log scale: further right is slower. The number is the median.
2005001k2k5kEvomi: median 779 ms, 90% < 1760 ms, 199/200Evomi779SpaceProxy: median 233 ms, 90% < 245 ms, 200/200SpaceProxy233
2005001k2k5kEvomi: median 779 ms, 90% < 1760 ms, 199/200Evomi779SpaceProxy: median 233 ms, 90% < 245 ms, 200/200SpaceProxy233
middle half of requests80% of requestsmedianms
ProviderSuccessResponse, msnew / openSpeedMbit/sBlocklistsZillowRedditTestedDataTrendResponse, ms
Evomi99.5%779 / 25511.15.5%62.5%62.5%September 25, 2026CSV
SpaceProxy100%233 / 20392.60%100%0%September 25, 2026CSVResponse, ms: 231 → 233

Individual IPv6 Proxies

Response time of every requestWhere the response times of 200 requests fall. Log scale: further right is slower. The number is the median.
2005001kSpaceProxy: median 351 ms, 90% < 364 ms, 50/50SpaceProxy351
2005001kSpaceProxy: median 351 ms, 90% < 364 ms, 50/50SpaceProxy351
middle half of requests80% of requestsmedianms
ProviderSuccessResponse, msnew / openSpeedMbit/sBlocklistsZillowRedditTestedData
SpaceProxy100%351 / 23749.8———September 25, 2026CSV

Dedicated Datacenter Proxies

Response time of every requestWhere the response times of 200 requests fall. Log scale: further right is slower. The number is the median.
1002005001kSpaceProxy: median 160 ms, 90% < 171 ms, 50/50SpaceProxy160
1002005001kSpaceProxy: median 160 ms, 90% < 171 ms, 50/50SpaceProxy160
middle half of requests80% of requestsmedianms
ProviderSuccessResponse, msnew / openSpeedMbit/sBlocklistsZillowRedditTestedData
SpaceProxy100%160 / 100182.10%100%0%September 25, 2026CSV

What the columns mean

How each number is calculated: formulas and test code on GitHub.

  • Success — share of requests that got an answer within 20 seconds.
  • Response, ms — median time per request: first on a new connection (a new IP each time), then on an open connection (sessions and static IPs). Measured from London, so compare providers with each other, not with tests from elsewhere.
  • Speed — download of a 2 MB file, transfer time only.
  • Blocklists — exit IPs found on FireHOL level 1-3 lists.
  • Datacenter IPs — for residential and mobile proxies, exit IPs that belong to hosting or cloud networks. Lower is better.
  • Zillow, Reddit — share of plain requests these sites let through. Both judge by IP reputation.

Each provider name links to its review with the full breakdown of the run. The full method is on our methodology page, the test code is open on GitHub, and the data is free to reuse under CC BY 4.0 with a link to this page.

Caproxy

Contacts

hello@caproxy.com

Follow us

Twitter Telegram Bluesky Medium Youtube Linkedin

More

Chrome Web Store Firefox Add-ons Google Play GitHub
© 2026 Caproxy

By type

  • Mobile Proxies
  • ISP Proxies
  • Datacenter Proxies
  • Individual IPv6 Proxies
  • Static Residential Proxies
  • Rotation Residential Proxies
  • Dedicated Mobile Proxies

By country

  • United States
  • Canada
  • Poland
  • Bulgaria
  • Croatia
  • Czech Republic
  • Denmark

By purpose

  • Multi-accounting
  • Web Scraping (data Parsing)
  • Antidetect Browser
  • Brand Protection

More

  • About Us
  • How we rate providers
  • Proxy benchmark
  • Advertising
  • Proxy Checker
  • Rating
  • Partners
  • Disclaimer
  • Privacy Policy

Languages

© 2026 Caproxy
  • Proxy selection
    • By type
    • By country
    • By purpose
  • Rating
  • Proxy Providers
  • Blog
  • For Companies
  • Proxy Checker
Rating List Blog