Apify is harder to price than most scraping vendors, and not because it hides anything. The pricing page is detailed. The trouble is that it bills resources, not pages: memory time, proxy gigabytes, storage operations and data transfer, all drawn from a monthly budget. To compare it with anything else, you have to turn those resources into a cost per page yourself.
That is what this piece does. Every rate below was read from Apify's pricing page on 21 September 2026, and the pages per compute unit figures come from Apify's own help center article on estimating usage. Vendors change prices, so check the live page before you commit a budget.
The plans
Every plan fee is also the amount of usage it includes. Pay 199 USD, get 199 USD to spend on compute, proxies and Store Actors, and anything past that is added to the next invoice. Annual billing takes 10 percent off.
| Plan | Per month | Compute unit | Residential proxy | Concurrent runs |
|---|---|---|---|---|
| Free | 0 USD (5 USD usage) | 0.20 USD | 8 USD per GB | 5 |
| Starter | 19 USD | 0.20 USD | 8 USD per GB | 32 |
| Scale | 199 USD | 0.16 USD | 7.50 USD per GB | 128 |
| Business | 999 USD | 0.13 USD | 7 USD per GB | 256 |
Datacenter addresses are bundled: 30 on Starter, 200 on Scale, 500 on Business, with extra IPs at 1, 0.80 and 0.60 USD each. Google SERP proxies are 2.50, 2 and 1.70 USD per 1,000 results pages, and the Unblocker is 1.50, 1.25 and 1 USD per 1,000 requests. Enterprise is custom. Startups can get 30 percent off Scale and students 30 percent off Starter and Scale.
What you actually pay for
A compute unit is one gigabyte of memory held for one hour. Run an Actor with 4 GB for 15 minutes and you have used 1 CU, whether it scraped 50 pages or 5,000. That is the core of Apify's model: you pay for the machine, and pages per machine hour is your problem.
On top of compute you pay for proxy traffic (by the gigabyte for residential), for storage reads and writes, and for data leaving the platform. Store Actors built by third parties can add their own charges, either per event such as a returned result, or per rental. Each Actor's page states which model it uses, and some pay per event Actors include platform usage while others bill it separately.
Apify's FAQ also says plainly that run cost depends on the number of retries when a request fails. A blocked request still used memory and still moved proxy bytes, so on hard targets you pay for the attempts that lost, not only the ones that returned a page.
Compute per 1,000 pages
Apify's help center gives two averages: about 300 pages per CU with a browser crawler, and about 3,000 pages per CU with Cheerio, which parses raw HTML without launching a browser. Heavy pages can take up to three times more. Using those averages:
| Crawler type | CU per 1,000 pages | Starter | Scale | Business |
|---|---|---|---|---|
| Headless browser | 3.33 | 0.67 USD | 0.53 USD | 0.43 USD |
| Cheerio (raw HTML) | 0.33 | 0.07 USD | 0.05 USD | 0.04 USD |
Those are low numbers, and for plain HTML on datacenter addresses they are genuinely hard to beat. If that describes your crawl, stop reading and buy Apify. The rest of this article is about what happens when it does not.
The proxy line is where the bill moves
Bundled datacenter IPs work on undefended sites. Retailers, marketplaces, travel and job boards mostly sit behind Cloudflare, DataDome or PerimeterX, which penalize datacenter ranges on sight, so a crawl of those targets needs residential traffic. That is billed by the gigabyte, and a gigabyte goes quickly.
The useful way to think about it: every 100 KB your crawler downloads per page adds about 0.80 USD per 1,000 pages on Starter. A raw HTML fetch of a product page is often a few hundred kilobytes. A full browser load pulls scripts, fonts and images too and commonly passes a megabyte unless you block those resources, which adds about 8 USD per 1,000 pages.
The Unblocker is the other route. At 1.50 USD per 1,000 requests on Starter it prices the hard part per request instead of per byte, which is easier to budget, and you still pay the compute of the Actor that calls it.
Cost per 1,000 pages, by scenario
Combining the two lines on the Starter plan, against a flat price per successful request. The page weights are illustrative, so replace them with what your own crawler logs.
| Scenario | Apify Starter | Flat per request |
|---|---|---|
| Plain HTML, Cheerio, datacenter IPs | about 0.07 USD | 0.98 to 0.33 USD |
| Browser, datacenter IPs | about 0.67 USD | 0.98 to 0.33 USD |
| Cheerio, residential, 250 KB per page | about 2.07 USD | 0.98 to 0.33 USD |
| Cheerio through the Unblocker | about 1.57 USD | 0.98 to 0.33 USD |
| Google results pages via SERP proxy | 2.50 USD plus compute | 0.98 to 0.33 USD |
| Browser, residential, 1 MB per page | about 8.67 USD | 0.98 to 0.33 USD |
The flat column is our own web scraping API pricing: 49 USD for 50,000 successful requests on Starter, 0.60 USD per 1,000 on Growth and 0.33 USD on Scale, with rendering, residential addresses and bypasses included and failed attempts not billed. We are an interested party, which is why the first two rows, where Apify is clearly cheaper, are at the top.
The breakeven
For an HTML crawl on residential traffic, the line sits at about 114 KB of transfer per page at entry level. Below that, Apify Starter is cheaper than 0.98 USD per 1,000. Above it, flat pricing is. At the upper tiers the line moves down to about 37 KB per page, comparing Apify Scale with our 0.33 USD Scale rate, and very few commercial pages are that light.
For a browser crawl on residential traffic there is effectively no breakeven. Compute alone is 0.67 USD, leaving about 0.31 USD for proxy bytes, which is under 40 KB per page. A rendered retail page never comes in that small.
Two things push the real number up from the table. Retries on defended sites use compute and bandwidth that returned nothing. And engineering time: somebody chooses the Actor, tunes memory, blocks images to save bandwidth and watches the success rate. That work is the actual product you are buying when you pay per page instead.
Where Apify is the right buy
- Large crawls of undefended HTML, where Cheerio on bundled datacenter IPs costs a few cents per 1,000 pages.
- A ready Store Actor already does exactly your job and its per result price works for you. Skipping the build has real value.
- You need the platform itself: scheduling, datasets, storage and integrations around many different scrapers your team writes.
- Small budgets. 19 USD is a lower entry point than most APIs, and it buys roughly 28,000 browser pages on datacenter IPs.
Where flat per request pricing wins
- Protected targets. Anything behind Cloudflare, DataDome or PerimeterX needs residential traffic and bypass work on most requests. The DataDome bypass API and the Cloudflare scraping API bill those pages at the same rate as a plain blog.
- Search results. One request returns one results page through the SERP API, under 1 USD per 1,000 instead of 2.50 USD plus compute.
- Budgets finance has to approve. A cost per page that does not move when a target changes its bot settings is easier to defend than a gigabyte meter.
- Teams without a scraping engineer. Nobody has to size memory or block fonts to keep the bill down.
If the scraped rows feed dashboards or a warehouse, the other half of the job is noticing when a source quietly stops delivering, and teams that monitor freshness and volume on those tables catch a blocked target in hours rather than at month end. That is true on any vendor, including us.
Questions buyers ask
How much does Apify cost per month? Starter is 19 USD, Scale 199 USD and Business 999 USD, with custom Enterprise above and 10 percent off annual billing. There is also a free tier with 5 USD of monthly usage and no card required. Each fee is prepaid usage, so the real monthly bill is whatever your runs consume, with overage added to the next invoice.
What is an Apify compute unit? One compute unit is one gigabyte of memory used for one hour. It costs 0.20 USD on Free and Starter, 0.16 USD on Scale and 0.13 USD on Business. Apify estimates about 300 pages per compute unit with a browser crawler and about 3,000 with Cheerio, so the crawler type changes the cost tenfold.
How much does Apify cost per 1,000 pages? On Starter, roughly 0.07 USD for plain HTML and 0.67 USD for browser pages in compute alone. Residential proxy traffic at 8 USD per GB is added on top, which lifts a protected, browser rendered page to several dollars per 1,000. The Unblocker, at 1.50 USD per 1,000 requests, is the per request alternative.
Does Apify charge for failed requests? Indirectly, yes. Apify bills the resources a run consumes, and its own FAQ says run cost depends on the number of retries when requests fail. Blocked attempts used memory and proxy bandwidth, so they appear on the bill. Success-only billing is a different model, so compare the two on the same target list.
Is Apify cheaper than ScraperAPI? For undefended HTML it usually is, since Cheerio compute on datacenter IPs costs cents per 1,000 pages. For protected or rendered pages it depends on page weight and retries. ScraperAPI's credit multipliers are laid out in ScraperAPI pricing per 1,000 pages, so you can put all three numbers side by side, and Bright Data pricing per 1,000 pages adds the per gigabyte model to the same table. Oxylabs pricing per 1,000 pages is the fourth model, a published rate card that changes with the domain you point at.
What is the cheaper alternative to Apify for protected sites? A flat per request API, once your pages need residential traffic. At 0.98 USD per 1,000 on our Starter plan and 0.33 USD on Scale, with bypasses and rendering included, the comparison turns in our favor above about 114 KB of residential transfer per page at entry level. Below that, and on undefended sites, Apify is the better deal.