Apify appeared in four of the eight searches we measured and Oxylabs in the TikTok one, which is how this pair got picked. What makes it worth a page is that the two vendors are unreadable in opposite directions, and you can see it in a single screenshot of each pricing page.
One publishes precise money for a unit you cannot predict. The other publishes a unit anyone can understand, attached to numbers that are floors rather than prices.
Apify: exact money, unforecastable unit
| Plan | Price/month | Usage included | CU rate |
|---|---|---|---|
| Free | $0 | $5 | $0.20 / CU |
| Starter | $19 | $19 | $0.20 / CU |
| Scale | $199 | $199 | $0.16 / CU |
| Business | $999 | $999 | $0.13 / CU |
Annual billing takes 10% off. The free tier needs no card and caps you at 16 GB of RAM and 5 concurrent runs. The store advertises 80,000+ Actors.
The unit, in their words: 1 CU = 1 GB of RAM running for 1 hour. And the expiry, also theirs: unused usage credits "are not rolled over… they expire at the end of the billing cycle."
Everything in that table is exact. What it cannot tell you is the cost of your job, because a compute unit depends on how much RAM the Actor's author allocated, how slow the target was that day, and whether a retry fired. None of those three is yours to set.
Oxylabs: clear unit, prices that are floors
Read from their pricing page today, quoted exactly as published.
| Product | As published |
|---|---|
| Residential proxies | "Starts from $6/GB" |
| Web Scraper API | "Starts from $49/mo" |
| Web Unblocker | "Starts from 9.4" |
| Mobile proxies | "Starts from 7.5" |
| ISP proxies | "Starts from 16" |
| SOCKS5 proxies | "Starts from 8" |
| Datacenter proxies | not shown |
A free trial is offered on several products; the length is not stated.
⚠️ Three of those rows have no currency and no unit
"Starts from 9.4" is not a price. We do not know whether it is dollars per GB,
dollars per 1,000 requests, or a monthly figure — and we are not going to guess,
because every plausible guess differs from the others by more than an order of
magnitude.
The two rows that are complete — $6/GB residential and $49/mo for the Web
Scraper API — are the only ones we would put in a budget, and even those say
"starts from". A floor with no published ceiling tells you what you cannot
spend less than. It does not tell you what you will spend.
We are recording this as an absence rather than filling it from a reseller page or a sales deck. A number we did not read at the source does not enter the table.
The two unknowns, side by side
| Apify | Oxylabs | |
|---|---|---|
| Money published | exact, four tiers | floors only, three without units |
| Unit published | exact: 1 GB RAM × 1 h | clear where shown: per GB, per month |
| Can you forecast one job? | no — CU depends on the Actor and the target | partly — per-GB forecasts if you know page weight |
| Entry | $0, with $5 of credit, no card | free trial, length not stated |
| Monthly floor | $19 for the cheapest paid tier | "from $49/mo" for the Scraper API |
| Idle month | credits expire, so you lose them | depends on the plan; not published here |
| Who maintains the scraper | whoever wrote that Actor | Oxylabs |
The asymmetry is worth naming: Apify's unknown shrinks after one real run, because you can measure the CU a job consumed and multiply. Oxylabs' unknown needs a salesperson, because the ceiling is not on the page.
Which is worse depends on whether you would rather run an experiment or have a conversation.
Where each one wins
Apify wins on starting. $0 with $5 of credit and no card, 80,000 Actors, and somebody has probably already written the thing you need. For a one-off job against an awkward target this is hard to beat, and the scheduling, datasets and exports are a product rather than an afterthought.
Oxylabs wins on the hardest targets and on enterprise terms. A residential pool that size with unblocking is what you buy when a site actively fights you, and when you need an SLA and an invoice rather than a credit card, "contact us" stops being a flaw and becomes the point.
Apify loses on expiring credit: a quiet month is money gone. Oxylabs loses on legibility — you cannot plan from that page alone.
Neither answers the question in the last section.
Where we fit, narrowly
HonestHook is neither a marketplace nor a proxy network. We sell read
access to twelve social platforms in one response shape, billed per API call,
with a published per-field contract: the same field in the
same place on all twelve, or null with a stated reason.
| Unit | Entry | Expiry | Scope | |
|---|---|---|---|---|
| Apify | compute unit | $0, $5 credit | credits expire monthly | marketplace, any target |
| Oxylabs | GB, or monthly | free trial, length unstated | not published | proxies, unblocking, scraper APIs |
| HonestHook | API call | free, 1,000 credits/mo | purchased never expire | 12 social platforms |
If your target is not one of those twelve, we have nothing for you and both of them do. That is most of the web, and pretending otherwise would make this page worthless.
The thing none of the three sells
An hourly archive of what is rising, kept and never deleted.
Apify and Oxylabs both answer what is this page now, at any scale. Only an archive answers what accelerated, and that cannot be bought retroactively at any per-GB rate — the past was not photographed.
Deeper on each: Apify · Oxylabs · if you are leaving Apify specifically, the four reasons people do · or the whole field.
Get a free key — 1,000 credits a month, no card. Start here.
Apify's four plan rows with monthly prices, included usage and per-compute-unit rates, the 10% annual discount, the 16 GB and 5-concurrent free-tier limits, the no-card terms, the "1 CU = 1 GB of RAM running for 1 hour" definition and the credit-expiry wording read from apify.com/pricing on 3 October 2026 — primary source. The 80,000+ Actor count read from apify.com/store the same day. Oxylabs' seven product rows are quoted exactly as published on oxylabs.io/pricing on 3 October 2026, including the three that carry no currency or unit and the datacenter row that showed no price; the free-trial lengths were not stated and are recorded as such. We did not substitute figures from any other source for the incomplete rows. Verify before deciding.