4G and 5G carrier proxies: the buying criteria that matter
Every mobile proxy listing says real 4G, and plenty now say 5G. Behind that one phrase sits a surprising range of things. A phone on a shelf with a SIM in it. A USB modem in a rack with forty others. A router with a carrier SIM on somebody’s balcony. Or a residential pool with a mobile label stuck on the product page because mobile sells for more. From the outside they all look like a hostname, a port and a password, and the listing rarely tells you which one you’re paying for.
I run mobile proxy farms in Singapore, modems and SIMs and the servers they hang off, so I see this market from the supply side. That’s also why I won’t rank unnamed providers here. I haven’t run a controlled test across the market, and a ranking built on vibes would be worse than none. What I can do is walk through the criteria that separate a real carrier proxy from a relabelled one, what each criterion does for account work, and how to check it yourself before your accounts depend on it.
What you are actually buying
Strip the marketing away and a mobile proxy is four things stacked together. A radio device, usually a USB modem or a phone. A SIM card on a specific carrier’s network. A piece of software that routes your traffic out through that device’s mobile data connection. And a slice of time on it, either all of it or a share.
The reason mobile proxies are worth more than datacenter ones comes almost entirely from the second item. Carriers put huge numbers of subscribers behind a small pool of public addresses using carrier-grade NAT, so any given mobile address is shared by many ordinary people at once. Platforms know this, and they’re cautious about punishing a mobile address because doing it cheaply would hit real customers. The older piece on why mobile proxies are the gold standard covers that mechanism in depth, so I won’t rebuild it here.
What matters for buying is this: the protection comes from the carrier, not from the word “mobile” on a product page. If the traffic doesn’t really leave through a carrier’s mobile network, none of that protection applies. So the first criterion, before speed or price or anything else, is proving the address belongs to a carrier.
Check the carrier ASN yourself
Every public IP address belongs to a network, and every network has an autonomous system number, an ASN, registered to the organisation that runs it. Looking up an address’s ASN takes a few seconds with any free IP lookup tool, and it’s the fastest single check you can make on a mobile proxy.
Connect through the proxy, open a page that shows your IP, and look up that address. The owning organisation should be a mobile carrier you recognise for the country you bought. In Singapore, for example, M1’s network is AS4773 and StarHub’s is AS9874. If the lookup comes back as a hosting company, a cloud provider or a small network nobody has heard of, you aren’t on a carrier connection, whatever the listing says.
A few subtleties. Some carriers run separate ASNs for fixed broadband and mobile, or announce mobile ranges from a parent company’s network, so a carrier’s name alone doesn’t always prove the range is mobile. Some lookup tools also label connection type, mobile versus residential versus hosting, and that label is worth reading, though it comes from databases that can be out of date. And do the check more than once. Rotate a few times and look up each address. They should all come back to the same carrier. A proxy that alternates between a carrier and something else is a pool with some mobile in it, which is a different product.
This check takes five minutes and filters out a meaningful share of listings on its own.
Dedicated modem or shared pool
Once you know the traffic really leaves through a carrier, the next question is whose traffic it is.
A dedicated mobile proxy means one modem and one SIM serving only you for the length of your rental. Nobody else’s sessions go through it. When the address rotates, only your activity has ever touched the addresses it’s handed. A shared mobile proxy puts several customers on one modem, or draws each request from a pool of modems that everyone on the plan shares.
For account work that difference is large. On a shared modem, the address your account uses right now may have been used a minute ago by another customer doing something you’d never do, and platforms see that history. It’s the same dirty-neighbour problem the residential comparison piece describes, concentrated onto fewer addresses. Shared plans are cheaper and they’re fine for things like checking ads or pulling public data. For accounts you care about, dedicated is the thing to pay for.
The catch is that providers use these words loosely. Ask directly: is this modem assigned only to me for the whole rental, and will anyone else’s traffic ever pass through it? Then watch for evidence. If your throughput drops sharply at particular hours, or you see rotations you didn’t trigger, someone else may be sharing your device.
Who controls rotation
Rotation is how a mobile proxy changes its public address, usually by forcing the modem to drop and re-establish its data connection so the carrier hands it a new one. How much control you get over that matters a great deal and is rarely described clearly.
The best setup gives you a rotation link or API call that you trigger yourself, plus an optional timer you can set or switch off entirely. You decide when an address changes. That matters because account work mostly wants stability. An account logged in through one address shouldn’t see it change mid-session, and the ideal is often a sticky address for the life of a session, with a rotation only between sessions or between accounts.
Weaker setups rotate on a fixed schedule you can’t change, or rotate on every request, or rotate whenever the provider’s system decides the modem needs a reset. Each of those can put an account through an address change at an awkward moment.
Things to test in a trial: trigger a rotation and time how long until the proxy passes traffic again. Confirm the address actually changed, because some rotations hand back the same one. Check whether a rotation you trigger ever affects anything besides your own port. And ask what happens if the rotation fails, whether the modem gets retried automatically or sits there dead until someone notices.
How long an address really holds
Related to rotation but worth its own check: when you don’t rotate, how long does the address stay put?
On a carrier network the public address isn’t permanently tied to the modem. The carrier can reassign it when the data session drops, when the modem moves between cells, or on its own internal schedule. How often that happens varies by carrier and by network conditions, and I’m not going to give you a figure, because it isn’t a constant and anyone quoting one as a spec is guessing.
What you can do is measure your own. During a trial, set the proxy to never rotate on its timer and log the exit address every few minutes for a day. You’ll learn how stable that particular carrier and modem are in practice. If the address changes several times an hour without you touching it, that’s a problem for logged-in work, whether the cause is the carrier, a weak signal, or a provider reset script you weren’t told about.
Ask the provider what their modems do on signal loss. A good answer explains how the device reconnects and whether you get told. A vague answer usually means the address will move at times you can’t predict.
What the 5G label means
5G sells. It’s also the label I’d scrutinise most.
A proxy is only 5G if the device it runs on supports 5G, the SIM’s plan allows it, and the modem sits somewhere with 5G coverage it can actually hold. A lot of hardware sold for proxy farms is 4G-only. Plenty of phones that are 5G-capable in the marketing are only capable on certain bands, or fall back to 4G indoors, in a rack, surrounded by other radios. So a listing that says 4G/5G often means “4G, possibly 5G sometimes”.
Ask which device model runs the proxy, and whether the provider can show you the current network type the modem is attached to. Some dashboards show this directly. If they can’t tell you, treat it as 4G.
Here’s the part the marketing won’t say. For most account work, 5G versus 4G barely matters. What you need from a carrier proxy is the carrier’s address space and a stable session. Speed helps with video uploads or heavy browsing, and 4G is already fast enough for logging in, posting and scrolling. Paying a premium for a 5G label only makes sense when you’ve confirmed you’ll actually get 5G and you have a bandwidth-heavy use for it.
Phone or modem: what the hardware changes
Mobile proxy farms run on two broad kinds of hardware, and buyers often assume one is better without knowing what the difference actually touches.
A USB modem is a small radio with a SIM slot, plugged into a server that handles the routing. A phone-based proxy runs the same job on a handset, sometimes with an app doing the routing, sometimes tethered to a computer. Both exit through the carrier the same way, and the carrier hands both the same kind of address. From the address alone, a platform can’t tell them apart.
What a proxy never does, whichever hardware it runs on, is make your browser look like a phone. The platform still sees your browser’s fingerprint, your user agent, your screen, your fonts. A desktop browser profile on a mobile address is a perfectly ordinary combination, because people tether laptops to phones all the time. But nothing about the proxy hardware adds mobile device signals to your session, and a listing that implies otherwise is overselling.
There’s one layer where hardware can leak through. The operating system that finally sends your packets onto the network leaves small traces in how those packets are built, things like default time-to-live values and the order of certain connection options. Passive fingerprinting tools can read those traces and guess the sending system. On most setups the packets that reach the carrier are built by the proxy server or the device doing the routing, not by your own computer, so what’s visible is the provider’s stack. Whether any given platform weighs this is not something I can tell you, and I’d be suspicious of anyone who claims to know for every platform. It’s a reasonable question to ask a provider, and a good one will at least understand it.
For most buyers, the practical takeaway is simpler. Pick on carrier, dedication and rotation control. The hardware matters mainly for reliability: well-cooled modems in a managed rack tend to stay up, and phones left charging on a shelf for months tend to develop battery and heat problems.
Location: the country, the city, the carrier
Location on a mobile proxy works differently from residential, and buyers often expect more precision than exists.
Country is reliable. A SIM on a Singapore carrier exits with a Singapore address. City is shakier. Carriers route mobile traffic through a small number of core network sites, and IP geolocation databases often place a carrier’s whole mobile range at one of those sites or at the capital. So a proxy sold as one specific city may look like that city in one lookup tool and somewhere else in another, and that’s normal behaviour for carrier ranges rather than evidence of a fake.
Carrier choice is the part people underrate. In most countries there are two to four major carriers, and platforms see their ranges differently depending on how many of their real users sit on each. If you run several accounts in one country, spreading them across more than one carrier gives you more independent address pools. Ask whether you can choose the carrier, and whether the provider’s stock on each carrier is real or the dropdown always lands on the same one.
Also check the rest of the story matches. A carrier address in one country with a profile set to another country’s timezone and language contradicts itself, and that’s on you, not the provider. The locale signals guide covers that side.
Protocols, UDP handling, IPv6
The protocol and what the proxy passes through affect whether a browser or app behaves normally over it.
Most mobile proxies offer HTTP and SOCKS5. Both carry ordinary web traffic fine. The difference shows up with UDP, which some apps and browser features use for calls, media and newer transport protocols. SOCKS5 can carry UDP if the provider supports it; HTTP proxies can’t. If an app expects UDP and can’t get it, it’ll usually fall back to TCP, which works but looks slightly different from a normal phone on a normal mobile connection.
IPv6 is the other one to ask about. Many carriers now give mobile devices an IPv6 address alongside IPv4. A real phone on that carrier would use both. A proxy might pass IPv6, block it, or ignore it. The risky case is when the proxy handles IPv4 but your own machine’s IPv6 escapes around it, which is a leak worth testing for separately. Ask the provider how they handle IPv6 on each modem. Then test from your side with a leak checker, because only you can see whether your setup is leaking your own address.
Bandwidth, speed, honest testing
Speed claims on proxy listings are close to meaningless without context, and I’d ignore any single number.
Mobile throughput depends on the carrier, the cell, time of day, how many other devices share the tower, the modem’s signal, and on a shared plan, how many other customers share the modem. A figure from a quiet hour on a good day tells you nothing about Tuesday evening. What you want to know is what you’ll get at the hours you actually work.
Two practical questions matter more than peak speed. Is there a data cap, and what happens when you hit it: does the proxy stop, slow down, or bill you extra? And is there a fair-use clause that isn’t on the pricing page? Ask for the exact terms in writing.
Then test. Run your real workload through the proxy at your real hours for a few days. Load the pages your accounts actually load. Watch for timeouts and stalls more than raw speed. A proxy that averages lower throughput but never stalls is better for account work than a fast one that drops a session twice a day.
When a modem dies
Modems and SIMs fail. SIMs get deactivated, modems overheat, USB hubs lose power, carriers have outages. A provider’s quality shows most clearly in what happens when your device stops working.
Ask how you’ll know. Some providers show modem status in a dashboard. Some send alerts. Some leave you to discover it when a session fails. Then ask what happens next. Do they swap you onto a new modem automatically? Is that new modem on the same carrier? Will your port and credentials stay the same, or will you have to reconfigure every profile that used it?
Support hours matter here too. If your accounts work in one timezone and the provider’s support is only awake in another, a dead modem at your morning can mean a lost day. Ask what support channel they use and when it’s staffed, then try it during the trial with a real question and see how long an answer takes.
And ask about replacement policy for addresses that turn out to be flagged. Carrier addresses rotate through many users, so it’s possible to be handed one that’s already under suspicion somewhere. A provider who’ll rotate or swap you on request is easier to work with than one who argues.
The acceptable-use conversation
A good mobile proxy provider will ask what you’re using the proxy for. Some buyers find that annoying. I’d treat it as a good sign.
Providers who run real carrier infrastructure depend on their SIMs staying active and their carrier relationships staying intact. If customers use modems for fraud, spam or attacks, the provider loses SIMs, and at worst loses the ability to operate. So a provider that asks questions, has a written acceptable use policy, and refuses obviously abusive traffic is protecting the infrastructure you’re renting. A provider that asks nothing and accepts anything is also accepting the customers whose traffic will make those addresses worse.
From your side, read the policy. Make sure your use fits it. Legitimate multi-account work, running several accounts for your business or your clients, managing ads, testing how a site appears from a location, fits comfortably inside most policies. Anything built on stolen accounts, payment fraud or deceiving a platform’s users doesn’t, and no proxy choice changes that.
How many accounts one modem should carry
A dedicated modem gives you one carrier connection. How many accounts you put through it is your decision, and it’s the one that most often undoes the value of buying a good proxy.
The protective crowd on a mobile address is made of other people. Your accounts benefit from sharing an address with ordinary subscribers. They don’t benefit from sharing it with each other. If ten of your accounts all log in through one modem, each from its own fingerprinted profile, the platform can still see ten accounts that appear on the same addresses at the same times, rotate together, and go quiet together. That pattern belongs to one operator, not to ten unrelated people who happen to share a carrier.
There’s no safe number I can give you, and the older piece on how many accounts per proxy explains why it depends so much on the platform and on how the accounts behave. The principle holds, though: the fewer accounts per modem, the less the modem’s address history ties them together. Accounts that must never be linked, like a client’s account and your own, shouldn’t share a modem at all.
Rotation doesn’t solve this, by the way. If several accounts ride the same modem, they get the same new address at the same moment when it rotates. Rotation changes which address they share. They keep sharing one.
When you’re budgeting, work backwards from this. Decide how many accounts need their own network identity, then buy modems to match, rather than buying one modem and finding out later how many accounts it can carry.
Using a trial properly
A trial is only useful if you use it to test the criteria above rather than to casually browse.
On day one, run the ASN check across several rotations, confirm the carrier, and check the connection type label. Trigger rotations and time them. Set the proxy to never rotate and start logging the exit address. Run a leak check for WebRTC and IPv6 from your actual browser setup.
Over the next days, run your real workload at your real hours and note stalls and timeouts. Look at the address log for unexpected changes. Ask support one real question and time the answer. If the provider claims dedication, look for any sign of other traffic.
Only then put one low-value account through it, one you can afford to lose, and let it do ordinary work for a while. The one-day provider evaluation guide goes through the day-one part in more detail. The point I’d add for mobile specifically is the address log. Carrier networks move addresses on their own, and a day of logging tells you more about session stability than any listing will.
Pricing shape without the sticker numbers
I’m not going to quote prices, because they vary by country, carrier and time, and any figure here would be stale in months. What’s more useful is the shape.
Dedicated mobile proxies are usually priced per modem per period, a day, a week, a month, with bandwidth either unlimited or capped. Shared mobile plans are often priced per gigabyte, which makes them look cheap per unit and expensive if your accounts load heavy pages. A dedicated modem costs more per month because the provider is tying up real hardware and a real SIM plan for you alone.
Two things to watch. A dedicated mobile proxy priced far below what the hardware and SIM plan could plausibly cost is a sign that it isn’t really dedicated, or isn’t really mobile. And a per-gigabyte plan that looks cheap can end up costing more than a dedicated modem once you count the real traffic of logged-in browsing. The piece on what a month on a cheap proxy plan really costs walks through that arithmetic.
Price the account, not the proxy. If one dedicated modem serves a handful of accounts that each matter to your business, the per-account cost is what you’re comparing.
Planning your exit before you start
The last criterion is the one nobody thinks about when buying: what happens when you stop.
Accounts that have logged in through a modem’s addresses for months have a network history tied to that carrier and that range. If you cancel, or the provider shuts down, or the modem is reassigned, those accounts suddenly appear from somewhere new. One change like that is ordinary, since people change carriers and move house. A whole group of accounts changing network on the same day is a pattern.
So ask a few questions up front. Can you keep the same modem and carrier as long as you keep paying, or does the provider reshuffle hardware between customers? If a modem is retired, will you be moved to one on the same carrier? How much notice will you get if the provider changes carriers or stops serving a country? And if you want to leave, can you run old and new setups side by side for a while so accounts can move gradually?
The guide on moving an account to a new setup covers the account side. On the buying side, the point is to prefer a provider whose answers give you control over timing. An exit you can pace across weeks is survivable. One forced on your whole fleet in an afternoon is the kind of event that shows up later in a ban review.
Red flags worth walking away from
Some signals are enough to end an evaluation.
An ASN lookup that returns a hosting company. A provider who won’t say what device or carrier sits behind the proxy. “Dedicated” with no clear answer about whether other customers ever touch the modem. Rotation that happens on the provider’s schedule with no way to control it or switch it off. A 5G label on hardware that can’t do 5G, or no way to see the network type at all.
No written terms on data caps or fair use. Support that doesn’t answer during your working hours. No stated policy on what happens when a modem dies. And the opposite of the acceptable-use point: a provider that openly advertises for uses built on fraud, because their addresses will carry that history.
None of these alone proves a provider is bad. Several together usually mean the listing is selling a label.
The honest limit
A good carrier proxy gives your accounts the address space ordinary people use, and a stable session through it. That’s a real and specific advantage. It doesn’t make an account safe on its own. Platforms weigh the fingerprint, the behaviour and the links between accounts alongside the address, and the tolerance carriers get from platforms has limits. A single mobile address with dozens of accounts behind it, all doing the same thing, is still visibly one operator.
The criteria above help you buy the right thing. They can’t tell you in advance how a particular platform will treat a particular carrier next month. That’s something you find out on accounts you can afford to lose, and then keep watching.
I run Singapore mobile proxy farms myself, real modems on real carrier SIMs, dedicated per customer, so these are the questions I expect buyers to ask me and the ones I’d ask anyone else. For the rest of the proxy guides and the setups I actually use, start on the Multi Account Ops home page.
Get new guides and videos first — join the Telegram channel.