Residential proxy pricing: what a GB buys for your job

Residential proxy pricing on proxymint is per GB, and the rate drops as the package grows. Each package is a one-off purchase with no renewal. What that buys depends on page weight: tens of thousands of HTML pages per GB, or a few hundred full browser renders.

A buyer with a scraping job does not budget in gigabytes. They budget in pages, products or searches, and the gigabyte only means something once the page weight is known. So this guide turns a GB into pages for three fetch modes, and says which jobs another tier fits better.

How residential proxy pricing works

The meter is traffic. A residential proxy routes each request through a home internet connection. The pool is shared across customers and the exits change, so the fair unit is the bytes a customer moves, not the number of IPs touched. A thousand small JSON calls and one heavy product page cost the same if they move the same bytes.

The rate falls with package size. Residential proxy cost per GB drops as the package grows, in steps rather than on a curve, and the whole package is priced at the rate for its size. Every step is on the pricing page.

The package is one-off. On proxymint, rotating residential proxies are sold in whole GB, paid once. Nothing renews and nothing is charged again until someone orders another package.

Targeting and rotation are not priced. Country, region and city targeting, and the rotation mode (a new IP every request, a 5 to 60 minute timer, or sticky while the device stays online), draw on the same balance at the same rate. HTTP and SOCKS5 are both included. The proxymint price depends on the quantity alone.

Residential and datacenter proxies are billed in different units because of where the traffic runs. Datacenter addresses sit in hosting networks and are rented per port. Residential traffic is carried over home connections whose exits change and drop off, so it is metered by the byte, and the home ISP address is what a strict target accepts. The two tiers side by side are covered in datacenter vs residential proxies.

Per GB, per IP and subscription models

Residential proxy plans come in three shapes across the market:

• One-off per-GB packages. Pay for a fixed amount of traffic, use it when the job runs. This is how proxymint sells residential, and the model behind pay as you go proxies.
• Monthly subscriptions. A monthly allowance of GB that renews on a date. A fit for steady usage; wasteful when usage swings.
• Per-IP pricing. A fixed address rented by the month. That is a static ISP product, not a rotating pool.

A subscription only pays off when its allowance is used: a plan used at 40% of its allowance pays for 2.5 GB for every GB the job moves.

What a GB buys in pages

A gigabyte is only a budget once it is converted into the unit the job counts, and the conversion depends almost entirely on how the client fetches the page.

Two starting points come from primary data. The HTTP Archive Web Almanac 2025 page weight chapter (published January 15, 2026) puts the median home-page HTML at 22 KB and the median mobile inner page, with every image, script and font, at 1.8 MB as of July 2025. A JSON endpoint is usually far lighter than either, but its size varies by site, so the figure below is an illustrative 5 KB, not a measurement.

Same gigabyte, roughly 80 times more pages when the client fetches HTML instead of rendering the page. The fetch method moves residential proxy cost far more than the package size does.

Traffic per 1,000 pages

The unit a scraping budget really uses is traffic per 1,000 pages. Because 1,000 pages at 1 MB each is 1 GB, the traffic for 1,000 pages in GB equals the page weight in MB. As one labelled scenario, full renders at the 1.8 MB median move about 1.8 GB per 1,000 pages; the table below stays in traffic, so the rate for any package size applies to it directly.

These figures assume every request succeeds. A block still moves bytes, so divide the traffic by successful pages, not by requests sent: at 80% success a full-render job pays for a quarter more traffic per page it keeps, unless the block page is small.

Add 20-30% for retries and heavier pages, and the 100,000-page render job moves to about 216-234 GB. Price that quantity on the pricing page; the HTML version of the same job fits in 3 GB.

Medians are a starting point, not the target’s real weight. A better estimate takes ten minutes:

1. Pick 20 pages that represent the job: listings, detail pages, search results.
2. Fetch them through the same client, headers and render settings the job will use.
3. Sum the bytes on the wire, or read the usage counter before and after.
4. Divide by 20, multiply by the page count, and add 20-30% headroom.

Then cut the bytes at the client. Send Accept-Encoding with gzip or br, defined in RFC 9110 section 12.5.3 (June 2022), fetch HTML or the page’s own JSON instead of rendering where the site’s terms allow it, and block images, media and fonts when a render is unavoidable. A blocked request retried ten times is ten pages of traffic and no data, so treat a block as a reason to stop.

Residential proxy price bands

Residential proxy pricing on proxymint is set in volume bands that step down as the package grows. The pricing page lists every band, and checkout charges against that table. The whole package takes the rate for its size, so the order is one number: the quantity, sized from the page estimate plus headroom.

Expiry and run-outs

Expiry. Across the market, some providers keep traffic on the balance and others expire it after a set period or at a subscription renewal. proxymint does not set an expiry on residential GB, so a package bought for a quarterly job is still there for the next run.

Running out. When the balance reaches zero, traffic stops. Nothing renews by itself and nothing is charged by itself; a new package is a new order. A scheduled overnight job needs its top-up before it starts, not after it fails halfway through a crawl.

Before buying cheap residential proxies

In residential proxy pricing, a low headline rate per GB is only one line of the bill. Before comparing cheap residential proxies, get answers to these, in writing, from the provider’s own terms page:

• Minimum package or minimum spend. A low rate that only applies above a large commitment is not the entry price.
• Expiry. How long unused GB last, and whether a plan’s allowance resets at renewal.
• Commitment. Monthly, quarterly or annual, and whether it renews automatically.
• Targeting surcharges. Whether city, region or session control is priced separately.
• Allowance used. How much of last month’s plan allowance the job really moved.
• Refund window. When it lapses and how much usage voids it.

A provider that answers all six clearly is easier to budget than one with a lower headline and vague terms.

When to buy something other than residential

Residential per GB suits targets that block datacenter ranges and jobs that need many consumer addresses. Three common jobs fit another tier better. How the per-GB, per-IP and per-port units compare across tiers is covered in proxy pricing.

Lenient bulk collection. API polling, uptime checks and targets without anti-bot scoring do not need a home ISP address. Datacenter proxies are sold per port per month, with no bandwidth cap. A port is billed the same at 2 GB or 200 GB, so the 180 GB render job above fits a handful of ports if the target serves datacenter addresses.

One address for weeks. Logged-in dashboards and partner systems that expect a fixed IP want the same exit every day. Static ISP proxies are billed per IP per month, each an ISP-registered address assigned to your order for the whole billing term, and a rotating pool cannot promise one address for that long.

Targets that only trust a mobile origin. Mobile proxies exit from mobile carrier networks, classified as cellular by MaxMind, which app-first platforms and mobile ad checks expect; price and catalog monitoring clears on residential. How that tier is billed is covered in mobile proxy pricing.

Which jobs fit residential per-GB pricing

Residential proxy pricing only settles into a budget when it is worked backwards from the job:

• Testing a target: a 1 GB package. Measure bytes per page on 20 real pages first.
• HTML or JSON collection: a small package goes a long way; 100,000 HTML pages fit in 3 GB.
• Full browser renders: size from the measured page weight plus 20-30% headroom, and block media before sizing.
• Steady monthly volume: buy one package sized to the month rather than several small ones.
• Lenient targets or one fixed address: datacenter ports or static ISP per month.