ZenRows advertises credits, and a credit is not a request. A plain request costs one credit, and a rendered request through premium proxies costs 25. A defended target usually needs both, so the pool is worth a twenty fifth of its face value. This page converts every published plan into money per thousand real requests, read on 9 September 2026.
We sell competing proxies and scraper endpoints, so weigh our reading accordingly. The multipliers below are theirs, and the division is ours.
The short version
The $57 Launch plan advertises 250,000 credits. Against a target needing JavaScript rendering and premium proxies, at 25 credits each, it holds 10,000 requests. That is $5.70 per thousand rather than the $0.23 the credit count suggests. The pricing page also states that you only spend credits on successful requests, and then states that 404 and 410 responses count as successful results.
What ZenRows publishes
These are the advertised plans, read from the pricing page on 9 September 2026. The fourth column is our own division, and it prices a credit rather than a request.
ZenRows plans and the price of one credit
| Plan | Per month | Credits | Per 1,000 credits | Concurrency |
|---|---|---|---|---|
| Free | $0 | 5,000 | free | 5 |
| Build | $16 | 45,000 | $0.36 | 20 |
| Launch | $57 | 250,000 | $0.23 | 50 |
| Growth | $165 | 1,200,000 | $0.14 | 100 |
| Scale | $456 | 5,000,000 | $0.09 | 200 |
| Enterprise | custom | unlimited | not published | 400 to 1,000+ |
zenrows.com pricing page, read 9 September 2026, with our own division
Those per-credit numbers look excellent, and they are also close to meaningless. Nobody buys credits, and the next table is the one that matters.
The multipliers, and what they do to the price
ZenRows publishes its credit costs plainly, which is more than several rivals manage. The published weights are short enough to quote in full.
Credits consumed per request
| Configuration | Credits |
|---|---|
| Plain request | 1 |
| JavaScript rendering | 5 |
| Premium proxies | 10 |
| JavaScript rendering and premium proxies | 25 |
| Residential bandwidth | 25,000 per GB |
| Browser session | bandwidth plus 5 per minute |
zenrows.com pricing page, read 9 September 2026
Multiply the credit price by the credit weight and the real rate appears. This is the table nobody publishes, so we built it:
Cost per 1,000 requests, by configuration and plan
| Configuration | Credits | Build $16 | Launch $57 | Growth $165 | Scale $456 |
|---|---|---|---|---|---|
| Plain request | 1 | $0.36 | $0.23 | $0.14 | $0.09 |
| JavaScript rendering | 5 | $1.78 | $1.14 | $0.69 | $0.46 |
| Premium proxies | 10 | $3.56 | $2.28 | $1.38 | $0.91 |
| Rendering and premium | 25 | $8.89 | $5.70 | $3.44 | $2.28 |
| Residential, per GB | 25,000 | $8.89 | $5.70 | $3.44 | $2.28 |
our own arithmetic on advertised figures, 9 September 2026
Read the first row against the last and the range is 25 to one. The same plan, the same vendor and the same month deliver rates that differ by a factor of 25.
Which row applies to you is not really a choice. Your target decides it, and defended targets are the reason people buy a scraping API.
What the plan allowance is actually worth
Restate the same arithmetic as request counts and the plan cards read very differently. Divide each credit allowance by 25, the weight for a rendered request behind premium proxies:
Advertised credits against defended requests
| Plan | Advertised credits | Defended requests |
|---|---|---|
| Free | 5,000 | 200 |
| Build | 45,000 | 1,800 |
| Launch | 250,000 | 10,000 |
| Growth | 1,200,000 | 48,000 |
| Scale | 5,000,000 | 200,000 |
our own arithmetic on advertised figures, 9 September 2026
The free tier gives 5,000 credits with no card, which sounds generous. Against a defended target it is 200 requests, which is a morning of testing.
We measured this multiplier behaviour across three vendors in why 250,000 credits is not 250,000 requests. ZenRows was one of the three, and the weights have not moved since August.
A realistic job, priced properly
A realistic job mixes those rows, so any honest estimate has to mix them too. Take 100,000 pages a month, where 70 percent need rendering and premium proxies.
That is 70,000 requests at 25 credits and 30,000 requests at one credit. The total comes to 1,780,000 credits, which sits between the Growth and Scale allowances.
Growth at $165 holds 1.2 million credits, so it falls short by 580,000. Scale at $456 holds 5 million, so it covers the job with 64 percent left unused.
Either way, 100,000 pages cost $456, which is $4.56 per thousand pages. That is fifty times the headline rate, because the headline prices credits and the job buys pages.
Pricing your own job takes three numbers and one division. Count the share of your targets needing rendering, premium proxies, or both together. Multiply each share by its credit weight to get credits per thousand pages. Then divide the plan price by its credit allowance and multiply the two together.
The successful request rule, read twice
One line on the pricing page answers the failure question. You only spend credits on successful requests.
The next line defines the term. Not Found (404) and Gone (410) responses count as successful results.
Both sentences are true together, and read in order they cancel most of the comfort of the first. A dead link on your list is a successful result, and it costs the full multiplier.
On a rendered request behind premium proxies, a 404 therefore costs 25 credits. At the Launch rate that is about half a cent for a page that returned nothing you wanted.
This matters most on lists you did not build yourself. Anyone scraping a sitemap of uncertain age should price the dead links in first. Only then is a comparison with a vendor that excludes them a fair one.
Residential bandwidth, priced in credits
Residential traffic is billed at a flat 25,000 credits per gigabyte. That is a credit figure, so the money it represents depends on which plan you hold.
Residential bandwidth, converted to money per GB
| Where you buy it | Published as | Per GB |
|---|---|---|
| ZenRows Build | 25,000 credits per GB | $8.89 |
| ZenRows Launch | 25,000 credits per GB | $5.70 |
| ZenRows Growth | 25,000 credits per GB | $3.44 |
| ZenRows Scale | 25,000 credits per GB | $2.28 |
| HProxy residential | priced per GB directly | $0.50/GB |
vendor pricing page and our own arithmetic, 9 September 2026
Our own gigabyte is priced as a gigabyte, from a balance that does not expire. The volume floor is $0.44/GB once you reach 2,000 GB+. The structural point is the unit rather than the rate, and it makes budgeting hard. Bandwidth whose price moves with your subscription tier cannot be modelled in advance.
What is metered on top
Browser sessions carry their own charge, at bandwidth plus a flat five credits per minute. A session held open therefore bills by the clock rather than by the request.
Concurrency is capped per plan rather than metered, from 5 on the free tier to 200 on Scale. That is a throughput ceiling, not a bill, but it decides whether a job finishes on time.
We found no separate charge for JSON output, for screenshots or for retries on the pricing page. The five weights in the credit table appear to be the whole story.
The minimum, the commitment and the expiry
There is no minimum spend and no commitment on the standard plans. The free tier is 5,000 credits a month with no card, refreshed monthly with no rollover.
Paying yearly is stated to give two months free, at pay for ten and get twelve. The page adds that a yearly plan releases the full year's credit allowance up front.
That last term is better than it sounds, and it is rare. A year of credits available immediately removes the monthly forfeit that most subscription scrapers impose.
The rollover wording covers top-ups specifically, which is a narrower promise than it reads. The page states that unused top-up credits roll into the next period. It does not say whether the monthly plan allowance behaves the same way.
Who this pricing suits
ZenRows suits high volume work against defended targets, where the 25 credit weight is unavoidable. At Scale that weight still lands at $2.28 per thousand requests, which is competitive. It suits anyone who wants the anti-bot work absorbed rather than assembled, which is a real product.
It suits low volume against defended targets poorly. On Build, the same request costs $8.89 per thousand, which is four times the Scale rate for identical work.
It suits stale URL lists poorly too, because 404 and 410 responses consume credits at the full multiplier. A list with 20 percent dead links raises the effective rate by a quarter.
Anyone comparing across this market should convert every vendor the same way. We did that for compute units in the Apify teardown and for page credits in the Firecrawl teardown. Search bundles get the same treatment in the SerpApi teardown.
What we could not verify
The pricing page does not publish the size or price of top-up credit packs. It states that unused top-up credits roll over, without saying what a top-up costs.
We could not confirm whether the monthly plan allowance rolls over like a purchased top-up. The page addresses top-ups and the free plan, and is silent on the ordinary case.
The Enterprise tier is listed as unlimited credits with custom pricing, so no rate exists. Concurrency there is given as 400 to 1,000 or more.
Prices and multipliers both move, and this vendor changes features often. Every figure here was read on 9 September 2026, so check the live page before committing.
Sources
- ZenRows, pricing page, read 9 September 2026. Plan names, monthly prices, credit allowances and concurrency limits.
- ZenRows, pricing page, read 9 September 2026. Credit weights of 1, 5, 10 and 25, and 25,000 credits per gigabyte of residential bandwidth.
- ZenRows, pricing page, read 9 September 2026. The statement that only successful requests spend credits, and that 404 and 410 responses count as successful results.
- ZenRows, pricing page, read 9 September 2026. Browser sessions at bandwidth plus five credits per minute.
- ZenRows, pricing page, read 9 September 2026. The yearly terms of two months free and the full year's allowance released up front.
- Our own August 2026 study of credit multipliers across three scraping APIs, which measured the same weights.
- Our own arithmetic on those published figures, 9 September 2026, shown in every table above.
