More unique IPs than any network on the planet, in the world's largest residential pool. That's Shifter.More unique IPs than any network, in the world's largest pool. That's Shifter.

See the numbers

Atlas

Proxy networks, measured in the open.

Independent tests of every major proxy network, scored and ranked with a published formula.

146,064,199 requests tested in the last 30 days
Shifter Atlas
Live
  • 1Shifter logoShifter94.1A+
  • 2Oxylabs logoOxylabs83.4A
  • 3Decodo logoDecodo81.5A
  • 4FleetProxy logoFleetProxy80.6A
  • 5Geonode logoGeonode80.3A
Score breakdownShifter
Reliability99.3
Speed71.2
Availability100.0
Coverage100.0
Diversity98.7
DC ASNs99.5
Authenticity100.0
Value88.9
Equal testing
No paid ranks
Open data
Custom weights
Live dashboard

The market, measured

Every figure below follows the time window you pick at the bottom of the screen.

Requests in the last 3 days
146,064,199
9.8/10 success181311.0M443.641 GB
Authentic
100.0%
Exit IPs in the requested country
Countries
236
Incidents
102
Mis-sold
0.9%
Median latency
613 ms
Success
9.8/10
Cheapest / GB
$0.270
Residential spread
9.3x
Leaderboard

Every network, ranked in the open

One published formula, the same tests for everyone. Re-weight it for your workload.

Network leaderboard

18 networks, openly ranked

#NetworkAtlas ScoreSuccessLatency p50Unique IPsDC ASNs$ / GBTrend
1Shifter logoShifterResidential94.1A+9.4/10#5486 ms+5.2%8.95M2210.6%#6$0.75#3
2Oxylabs logoOxylabsResidential83.4A9.9/10#2480 ms+9.3%7.17M2231.3%#15$2.50#11
3Decodo logoDecodoResidential81.5A10.0/10#1504 ms+8.8%7.03M2251.3%#13$2.50#13
4FleetProxy logoFleetProxyResidential80.6A6.8/10#12919 ms−0.8%7.15M2160.7%#10$1.85#6
5Geonode logoGeonodeResidential80.3A9.5/10#4810 ms−0.5%3.29M2230.0%#3$0.59#2
6DataImpulse logoDataImpulseResidential80.1A6.3/10#15471 ms+3.0%881K2040.0%#2$0.80#4
7ProxyEmpire logoProxyEmpireResidential79.9B+7.6/10#8799 ms−19.4%6.04M2190.5%#5$1.98#8
8Webshare logoWebshareResidential79.3B+8.8/10#6770 ms+12.0%3.76M2201.0%#12$2.00#7
9NodeMaven logoNodeMavenResidential77.7B+7.3/10#9512 ms+5.0%5.36M2210.6%#7$3.25#15
10SOAX logoSOAXResidential77.5B+3.1/10#16418 ms+6.2%1.24M2181.3%#14$2.20#10
11Proxy-Seller logoProxy-SellerResidential77.2B+6.4/10#14753 ms+6.7%3.78M2240.7%#9$2.10#9
12Byteful logoBytefulResidential75.6B+9.7/10#3517 ms+21.9%3.88M1331.3%#16$2.50#12
13Infatica logoInfaticaResidential74.6B+6.5/10#13709 ms+8.8%4.82M2250.8%#11$2.90#14
14Massive logoMassiveResidential74.2B+7.0/10#11346 ms+3.2%1.98M1340.1%#4$4.90#17
15Evomi logoEvomiResidential73.1B+8.6/10#7384 ms+21.4%867K2040.0%#1$3.98#16
16Rayobyte logoRayobyteResidential71.5B1.0/10#17888 ms+15.3%2.10M2112.4%#18$1.50#5
17Proxy-Cheap logoProxy-CheapResidential70.5B1.0/10#171.02 s562K2080.6%#8$0.78#1
18IPRoyal logoIPRoyalResidential66.4B7.2/10#101.08 s−6.2%2.89M2171.6%#17$5.25#18

Live tape

Live test feed · 34m ago

Methodology

How we test proxy networks

Every network gets the same requests, to the same 8 targets, from the same locations, on the same schedule. Nothing is self-reported: each figure on this page comes from traffic we sent and measured ourselves.

  1. 1 Unique IPs

    Census

    Counts the distinct exit IPs, cities and ASNs a network really hands out.

  2. 2 Success and latency

    Pulse

    Scheduled requests to every target, measured hour by hour.

  3. 3 Under pressure

    Load

    Sustained volume, to see if speed and success hold up.

  4. 4 Authenticity

    Audit

    Every exit checked for hosting ASNs, wrong geo and shared pools.

View full methodology

Atlas Score

What goes into the Atlas Score

One score from 0 to 100, built from eight categories. Each category is scored against the other networks in the same window, then weighted. These are the default weights; the leaderboard presets re-weight them for hard targets, cost, latency or compliance.

Default weights Balanced

100

points

8 categories, scored 0 to 100, then weighted

  • Reliability

    22%

    Share of requests that return a usable page, across every target, with retries counted.

  • Speed

    14%

    Median latency from request start to last byte, on normal requests.

  • Availability

    10%

    How much of the time the gateway answers at all, and how often incidents happen.

  • Coverage

    12%

    Countries and cities where the network delivers working exits, not just where it lists them.

  • Diversity

    8%

    Distinct IPs and ASNs observed per country, so a pool cannot win on a few thousand recycled addresses.

  • DC ASNs

    10%

    Penalty for residential exits that resolve to hosting providers.

  • Authenticity

    10%

    Exits that land in the country you asked for and behave like real consumer connections.

  • Value

    14%

    What a successful request really costs at list price, failures and retries included.

Grades From the final score
  • A+ 88+
  • A 80+
  • B+ 72+
  • B 64+
  • C+ 56+
  • C 48+
  • D below 48

Network stats

Every number a proxy network should publish

Providers publish pool sizes and country counts they choose themselves, and review sites repeat them. Atlas measures twelve figures for every network instead, from traffic we send ourselves.

IP pool

01

What a network really hands out.

  • Unique IPs by city
  • Live IPs by location
  • ASNs and ISPs

Performance

02

How it behaves on real targets.

  • Success per target
  • Latency p50 to p99
  • Uptime and incidents

Authenticity

03

Whether the pool is genuine.

  • Datacenter ASN share
  • Pool overlap
  • Geolocation accuracy

Coverage and cost

04

Where it works and what it costs.

  • Countries and cities
  • Cost per success
  • Rotation and sessions

Comparison

Compare proxy providers side by side

The top networks in Atlas on the figures that decide a purchase, over the last 30 days. Every value comes from our own measurements.

Compare proxy providers side by side
Metric
Shifter logo #1
Shifter
94.1 A+
Oxylabs logo #2
Oxylabs
83.4 A
Decodo logo #3
Decodo
81.5 A
Success
9.4/10
9.9/10
10.0/10 Best
Latency p50
486 ms
480 ms Best
504 ms
Unique IPs (30 days)
8.95M Best
7.17M
7.03M
Countries
221
223
225 Best
Cities
103632 Best
101067
100995
DC ASN share
0.6% Best
1.3%
1.3%
Price per GB
$0.75 Best
$2.50
$2.50
Network page See details See details See details

Buyer's guide

How to choose the best residential proxy provider

Six checks that separate a good proxy network from a good sales page. Atlas runs all of them for you, but they are worth knowing before any trial.

  1. Test on your own target

    01

    A network that almost never fails on an easy site can still fail one request in four on a search engine. Judge it on the target you care about, not the average.

  2. Count unique IPs, not pool size

    02

    Advertised pool size is a claim. What protects you from blocks is how many distinct IPs you actually receive, in the country and city you need, while you run.

  3. Check the ASNs

    03

    A residential pool should resolve to consumer ISPs across many ASNs. Hosting ASNs and a handful of carriers are a sign of a cheaper pool sold under a better label.

  4. Price results, not gigabytes

    04

    Divide the price by the requests that succeed. Retries cost bandwidth too, so a cheap network with a low success rate often turns out to be the most expensive one.

    Atlas guide Cost per success
  5. Watch the tail latency

    05

    Median latency hides the slow requests that stall a scraper. The p90 and p99 figures show how a network behaves on its bad minutes, when your jobs time out.

  6. Ask where the IPs come from

    06

    Legitimate residential IPs come from people who opted in and can opt out. If a provider cannot explain where its IPs come from, assume the worst.

    Shifter blog Ethical IP sourcing

Testing proxies yourself? Share your proxy speed, success rate and geolocation results. Contribute to Atlas →

Why Atlas exists

Setting the standard for transparency in the proxy industry

Shifter has operated residential proxy infrastructure since 2012, among the first companies in the world to do so. Much of what the industry now takes for granted began as our work. As the market has grown, so has the distance between what providers claim and what they deliver.

Atlas closes that gap. Every network, Shifter included, is tested with the same methodology, scored with the same published formula and reported in full. Buyers get verifiable data in place of marketing figures, and providers are judged on performance rather than promises.

Read more: How providers ethically source residential IPs

  • 24/7 Live testing On real traffic
  • 1 method For every network Shifter included
  • 0 Paid placements Or sponsored rankings
  • 100% Open data Free to download
FAQ

Frequently asked questions

Answers to common questions about Atlas, how proxy providers are ranked and what the numbers mean.

Atlas is an open, independent comparison of proxy networks. We send the same requests through every network we test and publish what comes back: success rate on real targets, latency, how many unique and live IPs each network has per country and city, how many ASNs it spans, how much of a residential pool is really hosting IPs, and what a successful request truly costs.

It depends on the job. A network that wins on a soft target can fall apart on an AI assistant or a search engine, and the cheapest price per GB is rarely the cheapest price per successful request. The leaderboard ranks networks on a balanced Atlas Score, and the presets above it re-rank them for hard targets, cost, latency or compliance, so you can see which network is best for your workload rather than in general.

The Atlas Score combines reliability, speed, availability, coverage, IP diversity, datacenter ASN exposure, pool authenticity and value into one number from 0 to 100. Each category is scored against the other networks in the same measurement window, then weighted. The default weighting is published on this page, and you can switch to another preset to see how the ranking changes.

No. There is no paid placement, no sponsored slot and no affiliate ordering. Every network is ranked by the same published formula on the same requests, and the full dataset can be downloaded so anyone can check the result.

Shifter goes through exactly the same tests, targets and formula as every other network, and is labelled on the page like any other entry. That is the point of publishing the method and the raw dataset: you do not have to take our word for the ranking, you can recompute it.

On easy targets most residential networks complete well over 95 percent of requests. The differences show on hard targets such as search engines, marketplaces and AI assistants, where success can drop below 80 percent. Always compare success rates on the target you actually care about, with retries counted, rather than one headline number.

Pool size is what a provider claims. Unique IPs are the distinct exit addresses we actually observe in a measurement window, per country and per city. A network can advertise tens of millions of IPs and still hand you the same few thousand addresses over and over, which is what gets requests blocked.

It is the share of exits in a residential pool that resolve to hosting providers instead of consumer internet service providers. Websites treat datacenter addresses as suspicious, so a residential pool with a high DC ASN share behaves more like a cheaper datacenter proxy than the product it is sold as.

We take each network's list price, apply the bandwidth our test requests actually used, and divide by the requests that succeeded. Failed requests and retries are paid for in real use, so they are paid for here too. That is why a network with a low price per GB can still be expensive per result.

The tests run continuously and every figure on the page is updated hourly. You can switch each one between the last 24 hours, 7 days and 30 days, and the live tape shows the latest test events across all 8 targets.

Yes. The full dataset behind this page is free to download as JSON, including every network's scores, windows and test events. You are welcome to use it in your own research or comparisons.

Ethically sourced residential IPs come from people who knowingly opt in, usually in exchange for a benefit inside an app, and who can opt out at any time. Networks built from malware, hidden SDKs or compromised devices are not legitimate residential proxies, and the authenticity and overlap tests in Atlas are designed to expose pools that behave like them.

Get started

Check the leaderboard against your own traffic

Rankings are a starting point. Order Shifter with a 3-day money-back guarantee and see how it holds up on the sites you actually scrape.

3-day money-back guarantee. Cancel anytime.