4G Rotating Proxy: How Rotation Works and Setup

A 4G rotating proxy routes your traffic through real phones on a carrier's LTE network and changes the exit IP on a schedule you choose: every request, every N seconds, or never (sticky) until you release the session. Because the exit is a carrier IP behind CGNAT, the address you get is shared with ordinary subscribers, which is why platforms treat it more leniently than a datacenter range. Setup is a connection string: host, port, credentials, and rotation expressed as parameters.

What rotation actually does

Rotation swaps which device carries your next request. Three modes cover every workload:

• Per-request: a new device, and therefore a new carrier IP, on every single request. Use it for independent, stateless requests (public product pages, SERP samples, price checks).
• Timed: one device for N seconds, then a swap. Use it for crawls that keep a cookie jar or a rate window alive for a few minutes.
• Sticky: one device held for the session, no automatic swap. Required for anything with a login, cart or multi-step form. See sticky sessions on a mobile proxy for how long to hold one.

Rotation is not anonymity. It spreads request volume across IPs so no single address hits a per-IP rate limit. Your browser fingerprint, TLS signature and headers travel unchanged, so a rotating IP behind a mismatched fingerprint still gets flagged.

4G vs residential vs datacenter

The reason a 4G IP survives where a datacenter IP does not is CGNAT: thousands of real subscribers sit behind the same public address, so a platform that bans it takes collateral damage. If you want the definition side by side, what is a residential proxy covers the ISP equivalent.

3G, LTE, 4G and 5G labels describe the radio link, not the IP quality. A 5G device delivers more throughput and lower latency; the IP still resolves to the same carrier ASN. Throughput only matters when your workload is heavy or concurrent, and 5G proxy speed benchmarks for TikTok automation has the measured numbers.

Setting up a 4G rotating proxy

Gateway (flex) mode

One credential, and you control geo and rotation per request by chaining underscore parameters onto the username. No _s_ parameter means per-request rotation:

curl -x "http://<username>_c_US:<password>@proxy.voidmob.com:10092" https://ipinfo.io/json

Run that twice and you should see two different IPs. To hold one device instead, add a session ID and a lifetime:

curl -x "http://<username>_c_US_s_run1_ttl_30m:<password>@proxy.voidmob.com:10092" https://ipinfo.io/json

Same _s_ value keeps the same exit IP; change the string to rotate on demand. Credentials for the gateway come from the flex credentials endpoint on your proxy order, so the snippet only works once the package is provisioned. Country codes are uppercase ISO (_c_US, not _c_us) and prefixes are case sensitive.

List mode

If you would rather configure rotation once than pass it per request, create a named proxy list with its own credentials, geo and rotation_period_seconds: 0 for per request, -1 for sticky, or N for every N seconds. The list returns ready-to-paste login:[email protected]:10000 entries, which is what most scrapers, bots and antidetect browsers want. Geo on a list is set by fields, not by username suffixes, so do not mix the two mechanisms in one list. Incogniton's proxy manager is a worked example of feeding list entries into a browser profile.

Both modes can be active on the same package and draw from the same GB balance, which is the usual pattern: lists for browser profiles, flex for code.

Pick the rotation interval from the workload

The interval is a function of how much state the target expects, not of how fast you want to go:

• Stateless GET requests: per request. One IP per page view keeps you under every per-IP rate limit.
• Paginated crawl with a session cookie: sticky for the crawl, then a new session ID for the next batch. A mid-pagination IP change often resets or invalidates the cursor.
• Login, checkout, posting: sticky for the whole flow. Rotating during a login looks like session hijacking.
• Long automation runs: 30 to 120 minutes on one device, then rotate. That is roughly how often a real phone changes address as it moves between towers.

The failure mode people hit first

Two configuration mistakes account for most "rotation is broken" tickets.

First, a parameter that controls behavior on node failure is not a rotation interval. It decides whether a dropped device is replaced instantly, after a short delay, or not at all. Setting it does not make the gateway rotate every N seconds; only the timed rotation setting does that.

Second, a sticky session that is not given an explicit lifetime expires after a period of inactivity. If your script pauses between steps longer than that, the pin releases and your next request exits from a new device, mid-flow. Set _ttl_ to cover the whole run, and re-check the exit IP before any step that depends on the previous one.

Sizing and cost

Rotating pools are billed by data, so cost is a function of bytes, not IPs. Shared GB packages start at $3.99 for 1 GB and scale down per GB as the package grows: $35 for 10 GB, $81.25 for 25 GB, $149.50 for 50 GB, and $250 for 100 GB, which works out to $2.50 per GB at the top tier. A dedicated device, where you own rotation outright and traffic is not metered per GB the same way, starts at $49 per month depending on country.

Estimate before you buy: HTML-only scraping is cheap per request, while a headless browser loading images and scripts can burn 2 to 5 MB per page.

If the workload is a handful of long-lived accounts rather than volume scraping, a dedicated device is usually cheaper and safer than a rotating pool. Dedicated vs shared mobile proxies lays out that decision, and USA rotating proxies covers the US-specific rotation rules.

Verify before you run volume

Three checks, in order:

1. Exit IP changes as configured. Two calls with no session ID should return two addresses; two calls with the same _s_ should return one.
2. The IP resolves as mobile. Run it through the IP Type Checker and confirm the carrier ASN, not a hosting range.
3. Auth and protocol work from your client. The Proxy Validator confirms credentials, reachability and anonymity level before your code does.

A pool that passes all three and still gets blocked is a fingerprint problem, not an IP problem.