SOAX vs Oxylabs: Pool Size, Pricing and Session Control

Neither network wins across the board. They win in different volume bands and on different capabilities, so the decision is usually your monthly gigabytes plus one spec.

If you pull 300 GB a month, all US, with sessions under ten minutes, stop reading and buy SOAX. If you pull 15 GB a month and hold logged-in sessions for hours, buy Oxylabs. The rest of this article is for everyone in between.

Every SOAX figure was read from SOAX's own pricing, residential proxies and mobile proxies pages on 8 September 2026. Every Oxylabs figure was read from Oxylabs' own pricing, residential proxy pool and datacenter proxies pages on the same date. Nothing here is a benchmark SparkProxy ran. Where a vendor publishes a success rate, a response time or a pool count, we label it as that vendor's own claim, because nobody audits those figures, ourselves included. Prices move with promotions and renegotiation, so check both current pages before you spend anything on the strength of a table in an article.

Disclosure: SparkProxy sells rotating datacenter proxies and a Scraping API, so we compete with part of what Oxylabs sells and none of what SOAX sells. Section 10 says when neither of these two is the right buy, including the case where we are not either. Wider roundups sit at SOAX alternatives and Oxylabs alternatives.

Both vendors lead with pool size. It should carry the least weight in your decision.

Those are vendor marketing claims, not audited counts, and the two success rates match to the decimal place. The pool gap is roughly 13 percent. No target's rate limiter can tell a 155 million pool from a 175 million one, because it never sees the pool. It sees the handful of exit addresses you route through it that day.

What decides your block rate is whether the subnets you get assigned are already burned on your target. A pool count cannot tell you that and neither can a review. Only a trial against your real URLs can, which is why section 11 exists. If pool composition is new to you, what is a proxy pool covers how these networks are assembled.

Here is SOAX's published plan ladder as of 8 September 2026. All plans exclude VAT, credits are prepaid, and one credit equals one dollar of usage spent at your plan's per-GB rate.

The Enterprise rate is itself tiered by consumption: $0.85/GB to 5 TB, $0.75/GB from 5 TB to 15 TB, $0.50/GB above 15 TB, with each gigabyte staying in the bucket it landed in. Unused credits roll over 60 days on monthly billing, 365 days on annual. Postpaid overage invoicing is published as unavailable on Sandbox, Builder, Team and Scale.

Oxylabs residential, published on the same date:

Read them together and the structural difference is the whole comparison. SOAX gates its low rates behind a plan fee: $1.50/GB requires committing $1,500 a month. Oxylabs gates its low rates behind bundle size: $2.50/GB requires buying a terabyte. Both sell volume commitment, they just dress it differently.

Verify one thing yourself before modelling anything. Confirm the credit allowance attached to the plan you are considering, straight from SOAX's current pricing page. It is the single number that decides this comparison, and a pricing table that also lists rates across three country tiers is easy to misread.

Rates are quotes. Plans are what you pay. This table takes a monthly Tier 1 volume and finds the cheapest published option on each side.

Four caveats. SOAX figures exclude VAT and assume Tier 1 exits. The SOAX rows above the included credit allowance assume you can top up prepaid credits at your plan rate, worth confirming given that postpaid invoicing is published as unavailable below Enterprise. SOAX credits expire on the rollover clock while an Oxylabs bundle is data you already own. And both vendors negotiate above their published tiers, so the 5 TB row starts a conversation rather than ending one.

The pattern in that table is not about either brand. It is about step size. Oxylabs' ladder goes 5, 20, 125, 1000, a 6.25x jump from Basic to Advanced. A team consuming 40 GB a month cannot buy 40 GB. It buys the 125 GB Advanced plan at $500 and uses under a third of it:

effective rate = plan price / GB actually consumed
$500 / 40 GB = $12.50 per GB consumed, against a $4 label

That gap is the most common way buyers overpay for residential proxies, and it has nothing to do with network quality. Compute your effective rate first, using gigabytes you will really consume rather than gigabytes included. The same arithmetic applied to other billing shapes is in understanding datacenter proxy pricing models.

Every head-to-head we have read on this query skips this, and it can swing the numbers above by a wide margin.

SOAX publishes its per-GB rates by country tier. The figures in this article are the Tier 1 column, covering geos such as the US, the UK and Japan, and Tier 2 and Tier 3 geos are published at lower rates. Oxylabs publishes a single global residential rate per plan on the pages we read, so a gigabyte from Germany and a gigabyte from Vietnam cost the same.

Your country mix is therefore an input to the SOAX price, not a detail. Three teams paying the same plan fee can land on very different blended rates. Before comparing, export a month of requests grouped by exit country, weight each group by bytes, and read the current rate for those countries off SOAX's pricing page rather than reusing a blended average somebody else calculated. If your collection is Tier 1 only, tiering gives you nothing and the Tier 1 column is the whole story. Geography also decides which network can serve your requests at all, the subject of regional vs global proxy pools.

Price bands change every quarter. These rows change your code.

Three rows in that table decide purchases on their own.

The sticky ceiling. Sixty minutes against twenty-four hours matters if your workflow holds an authenticated session, walks a multi-step checkout, or paginates a result set tied to one IP. On SOAX you design for re-authentication inside the hour. On Oxylabs you hold one exit for a working day. If your flow is stateless and rotates per request, ignore this row entirely. The trade-offs either way are in what is a sticky session proxy.

Transport. SOAX publishes UDP and QUIC support on both its residential and mobile products. The Oxylabs residential page we read lists neither. If you need to proxy HTTP/3 rather than force traffic down TLS over TCP, that is a capability difference, not a preference.

Targeting depth. Coordinates, ZIP and an OS filter exist on the Oxylabs side and are absent from SOAX's published list, which stops at city and ASN or ISP. Radius targeting around a point is hard to work around if you need it. City and ASN, which both offer, covers most commercial collection.

Expect an identity or use-case verification step on either network. That is normal for consumer IP supply, and the rest of the pre-purchase checklist is in what to evaluate when selecting a residential or datacenter proxy service.

The catalogs are not the same size, which matters if you want one vendor for everything.

SOAX's proxy menu is residential and mobile, alongside managed scraping products. As of 8 September 2026, soax.com/datacenter-proxies returns HTTP 404, so there is no datacenter or ISP tier to fall back on inside the account.

Oxylabs sells residential, mobile, ISP and datacenter proxies plus Web Unblocker and a Web Scraper API listed from $49 a month. Its datacenter product publishes 2M+ IPs across 188 locations, shared IPs from $0.70 each, dedicated IPs from $1.20 each, and a shared per-GB option from $0.44/GB. The fair usage terms are published plainly: 100 concurrent sessions per datacenter proxy until 50 GB of monthly usage on that proxy, then 10 sessions for the rest of the cycle. Publishing exact thresholds is more transparency than most of this category offers.

For an engineer who wants a block-rate number by Thursday, SOAX's $1.99 trial is the faster path on residential, and Oxylabs' 5 free datacenter IPs are faster if datacenter is what you are testing. For a procurement-led buy, the Oxylabs request flow attaches an account manager early, which is a feature rather than friction.

Buy SOAX when

• You consume 40 GB a month or more and want the middle of the range priced properly rather than rounded up to the next bundle.
• A meaningful share of your traffic exits outside Tier 1 countries.
• You need UDP or QUIC through the proxy.
• Your sessions are stateless or shorter than an hour.
• You want a self-serve paid trial today, with no sales conversation.

Buy Oxylabs when

• You buy under about 20 GB a month, where a $30 or $100 bundle beats a plan fee.
• You need sticky sessions measured in hours, not minutes.
• You need ZIP, coordinate, IP-version or OS-level targeting.
• You want residential, ISP, mobile and datacenter proxies plus an unblocker on one invoice.
• Procurement requires enterprise contracting and a documented fair usage policy.

Neither list contains "has the bigger pool", because at these sizes that criterion decides nothing.

Residential IPs are the most expensive way to make an HTTP request, and plenty of teams paying for them do not need them on every target. Consumer IP space solves one problem: looking like a household connection to a site that classifies address space. If your target does not classify by IP class, you are paying a premium for nothing.

Check that before you compare vendors. Public catalogs, filings archives, sitemaps and feeds, long-tail marketplaces, availability monitoring and localisation QA usually run fine from datacenter ranges. Social platforms, account-bound flows, mobile-gated surfaces and hard targets behind a top-tier bot manager do not, and no pricing argument changes that. The full split is in residential vs datacenter proxies, and the three-way version in comparison of residential vs datacenter vs mobile proxy types.

For traffic that does move, the billing unit changes entirely. SparkProxy sells rotating datacenter proxies on flat monthly plans with unlimited bandwidth and 30 days validity, billed by concurrent threads rather than gigabytes:

Those plans draw on 1M+ datacenter IPs across 80+ countries, including more than 50,000 US datacenter IPs, reached through gateway.sparkproxy.io on port 11000 for HTTP and HTTPS, 11002 for sticky sessions and 13000 for SOCKS5. Each tier carries a fair usage speed cap, from 25 Mbps on Starter to 150 Mbps on Plus, and those are ceilings rather than guaranteed rates. Targeting is country level only, with no city, ASN or ZIP granularity, so this does not replace what either vendor above sells.

Put that beside the metered tables. A 250 GB month costs $550 on SOAX Team and $2,500 on Oxylabs Corporate at list. The same workload on a flat datacenter plan costs $240 whether it moves 250 GB or 900 GB, and the comparison holds only if your target accepts datacenter IPs. Test that first, on your real URLs.

If what you are buying is anti-bot handling rather than geography, a managed endpoint may undercut both proxy bills, because failed retries stop being billable work you own. SparkProxy's Scraping API starts with 1,000 free credits and no card, then $49 for 250,000 credits a month at 50 concurrent requests, $99 for 1,000,000 at 100 concurrent, $249 for 3,000,000 at 200, and $599 for 8,000,000 at 400. A plain fetch costs 1 credit, a JavaScript render 5, a screenshot or PDF 10. That trade is worked through in web scraping API vs self-managed proxies, and the direct head-to-head with one of the vendors here is in SparkProxy vs Oxylabs.

Do not buy either network off a pricing page. Two trials and one number settle this.

1. Pick 1,000 real URLs from your target list, including the hardest page type you scrape and your true country mix.
2. Run all 1,000 through each trial with identical code, headers and rendering settings. Change only the proxy endpoint.
3. Record three numbers per vendor: requests that returned the data you wanted, total bytes transferred, median latency.
4. Convert to gigabytes per 1,000 useful records. That is your real consumption unit.
5. Project a full month at that consumption, then price it on both ladders using the plan you would actually land on.
6. Only now compare success rates, using your measured numbers rather than either vendor's published 99.95 percent.

Step 5 in code, using the two published shapes:

SOAX_PLANS = [("Sandbox", 0, 5.00), ("Builder", 200, 3.00), ("Team", 500, 2.20),
              ("Scale", 1500, 1.50), ("Enterprise", 3000, 0.85)]   # Tier 1, excl. VAT
OXY_BUNDLES = [(5, 30), (20, 100), (125, 500), (1000, 2500)]       # GB included, price

def soax_month(gb):
    """Plan fee is prepaid credits at 1 credit = $1, spent at the plan's per-GB rate."""
    return min(max(fee, gb * rate) for _, fee, rate in SOAX_PLANS)

def oxylabs_month(gb):
    """Smallest published bundle that covers the volume."""
    for included, price in OXY_BUNDLES:
        if gb <= included:
            return price
    return None      # above the published ladder: request a quote

for gb in (5, 20, 50, 100, 250, 500, 1000):
    print(gb, soax_month(gb), oxylabs_month(gb))

Feed in your own measured gigabytes, not a guess, and re-read both pricing pages before trusting the constants. A network three points better on success rate but twice the effective rate is not the better buy, and running the same URLs through both trials is the only way to tell which one you have.