BrightData, Oxylabs, and NetNut can be named as the most costly providers. In this group, BrightData is ahead by a wide margin regarding the number of unique addresses. NetNut has very few unique addresses. In terms of latency, Bright, and NetNut are both quite fast. Oxylabs latency has degraded significantly since Sprint 2024. IP quality is quite the same for Bright and NetNut; Oxylabs is step ahead here because of % of residential IPs.
Infatica, SOAX, and Geonode can be grouped in the middle based on pricing and performance. Infatica has significantly more IPs that two others, but they are intensively shared with other vendors - so the quality is bad. Geonode and SOAX are pretty strong regarding "truly unique"(not shared with anybody) IPs, but SOAX has significantly fewer total IPs.
From the technical Success Rate(tech SR) and latency perspective, SOAX, Infatica and Smartproxy are top3 overall.The quality of IPs also differs significantly. SOAX is at the top overall, but Geonode and Infatica is twice as bad and at the bottom of the list.
IPRoyal, Smartproxy, ZettaProxies. ZettaProxies and Smartproxy have two times more IPs than IProyal. But "truly unique." Smartproxy overperforms these two significantly. IProyal and Smartproxy have pretty good technical Success Rates (tech SR), and Smartproxy has pretty good latency at the same time. Talking about quality, SOAX and Smartproxy are leading because of the IPv6 share(I hope I'll have time to make IPv4 only comparison), IProyal is somewhere in the middle, but ZettaProxies, IPRoyal, Infatica has the low pool quality in this research. So, I can conclude that Smartproxy has top quality and performance and affordable prices simultaneously.
Methodology
Rarad chart: the higer - the better. Points and positions calculation is here.
The test load generator was placed in Europe/the Netherlands. It uses the HTTP CONNECT proxy protocol and makes 100 requests per second for the 30min. The GEO filter with Country=US were used for all vendors.
The test was repeated a few times on different days. Some tests were repeated using target resources that have IPv6 and those that don't.
Smartproxy, BrightData, NetNut were tested via resellers. So, the latency and SR may be slightly worse than with direct use.
This metric tells how many unique IP addresses in total you will get from the exact vendor using autorotation on each request feature doing 100 requests per second for a 30min. I need to highlight that next time I'll adjust the test because BrightData have the potential to return more unique IP addrs.
warning
Not all possible unique IPs were gathered during tests. Next time I'll increase the duration and/or RPS.
For example, BrightData returns 401K unique IPs for 60min / 200 rps test.
IPv4 + IPv6
IPv4 only
Target service has IPv4 and IPv6 addresses. So, residential devices might use it's IPv6 if possible.
This metric shos amount of unique IP addrs among all tested vendors. It differs from the previous one cause they have intersections in their IP addrs pools.
IPv4 + IPv6
IPv4 only
Target service has IPv4 and IPv6 addresses. So, residential devices might use it's IPv6 if possible.
We see that SOAX and IPRoyal maintained the tempo at the beginning and fell later. BrightData has a significant reserve of unique IP addresses. Infatica, ZettaProxies, and NetNut have remarkably similar statistics.
Unique for vendor - A number of unique(in the scope of exact vendor test) IPs were obtained from the exact vendor with the individual test.
Unique for all vendors - A number of unique IPs(in the scope of all conducted tests among all vendors) we got from the exact vendor with the individual test.
Share of unique in vendor(%) - Percentage of unique IPs in exact vendor return.
Share of unique in all(%) - Percentage of unique IPs among all tests in exact vendors return.
IPv4 + IPv6
IPv4 only
Target service has IPv4 and IPv6 addresses. So, residential devices might use it's IPv6 if possible.
Vendor
Unique for vendor
Unique for all vendors
Share of unique in vendor - %
Share of unique in all - %
BrightData
147992
86281
58
24.87
Geonode
78967
59735
76
17.22
Smartproxy
90740
51146
56
14.74
Oxylabs
116443
42650
37
12.29
SOAX
47098
34837
74
10.04
IPRoyal
49043
27829
57
8.02
Infatica
107701
17366
16
6.66
ZettaProxies
107743
13630
13
3.93
Netnut
107443
13455
13
3.88
Target service has only an IPv4 address. So residential devices don't have a chance to use their IPv6.
X-axis quantities and Y-axis values for time duration from the request start event til the moment when the proxy client gets the first byte of the response from the Target resource; in our case, this is a public IP checking service.
The event description can be found here Latency section.
The Netherlands-based data center performed all tests. Tool with the custom scheduler were built on top of ProxyChick.
warning
Be aware that latency for US and Asian proxy clients will be different.
Some vendors are tested via resellers frontends, which might create extra latency. If you are the vendor and want to give me an account for tests to get more precise results - let me know.
Resi IP Share - Percent of IPs that treated as residential(real devices that belong to real humans) by two popular IP DBs.
IPv6 if AAAA DNS records for target is available - percent of unique IPs that will have version 6 in case AAAA DNS recond available for target resource.
Clean IPs % - Percentage of unique IPs that aren't mentioned as Spam, Malicious, Proxy, etc, in popular data sources. One DataSource might be a false positive, so I created two metrics with and without it. 4 sources metric is more strict, 3 is less.
Thanks ip2location 🍦 for help with the data for this research ! With their help, I've calculated the quality using all gathered IPs instead of just the samples.
tip
If you have difficulties with bypassing antibot solutions and are not ready to develop your own solution - try using ready-made services like ScrapeOps
IPv4 + IPv6
IPv4 only
Target service has IPv4 and IPv6 addresses. So, residential devices might use it's IPv6 if possible.
Vendor
Clean IPs %(4 sources)
Clean IPs %(3 sources)
Resi IP Share - %
IPv6 if AAAA available - %
SOAX
69.4
84.0
98.2
57.77
Geonode
54.9
67.6
79.0
1.01
Oxylabs
40.8
64.6
92.0
16.14
Smartproxy
50.9
77.8
69.0
21.17
BrightData
32.9
59.3
93.1
0.03
Netnut
27.3
50.1
88.2
0.11
ZettaProxies
24.8
50.2
90.2
0.09
Infatica
21.7
50.4
90.1
0.16
IPRoyal
19.9
56.6
89.5
0.04
Target service has only an IPv4 address. So residential devices don't have a chance to use their IPv6.
Gathered from public pricing pages in November 2024
tip
If you like the features and performance of the mentioned proxy providers but are not comfortable with the prices or KYC, here are the alternative options ProxyEmpire