What 500,000 pages a month costs on 8 scraping APIs

Prices are the providers' published list prices, checked against their official pricing pages on Sep 25, 2026. Every product name below links to its page on ScrapeSelect, where each rate cites the page and excerpt it came from.

How we priced it

The workload is 500,000 pages a month that come back successfully, billed monthly. For each product, our cost engine takes the provider's own price table, works out what one page costs in that product's billing unit, adds retries for pages that fail, and picks the cheapest published plan that covers the month, including overage and top-ups. Temporary promotions are left out.

Some inputs can't be known before you run the scraper: how many requests fail, how many pages need premium proxies, and, for Zyte, which difficulty tier it assigns to your site. So every product gets three estimates. The low one uses optimistic values for those inputs, the high one pessimistic values, and the ranges in the tables come from that. The rest of the workload stays fixed: pages average 300 KB, and a rendered page takes 10 seconds of browser time. The methodology explains each assumption.

Seven of the nine products are managed scraping APIs and unlockers: you send a URL and get the page back. Two, Apify and Browserless, are platforms where you run your own scraper, so their price covers compute or browser time, not the scraping logic.

Easy: static pages, no bot protection

Plain HTTP requests, no JavaScript rendering, no country targeting.

Assumptions (low / expected / high): success rate 99% / 97% / 94%; pages needing premium proxies 0% / 0% / 10%; website difficulty tier 1 / 1 / 2. * You run your own scraping code on these platforms. Open this workload in the calculator to see today's prices or change the inputs.

• Zyte is cheapest at $65 because it prices by site difficulty, and an easy site sits in its lowest tier at $0.13 per 1,000 pages.
• ScrapingBee and ZenRows both land on $99 plans: a plain request costs one credit, so 500,000 pages fit. On ScrapingBee that means turning rendering off, since it is on by default and costs 5 credits.
• Oxylabs Web Unblocker is the most expensive because it bills by traffic. At 300 KB a page, 500,000 pages are about 150 GB a month. With much lighter pages it gets much cheaper.
• Bright Data charges a flat rate per successful request whatever the page needs, so for easy pages you pay for unblocking you don't use.

Typical: some JavaScript, medium protection, US IPs

20% of pages need JavaScript rendering, the site uses some bot protection, and traffic must come from the United States.

Assumptions (low / expected / high): success rate 97% / 93% / 88%; pages needing premium proxies 20% / 30% / 45%; website difficulty tier 2 / 3 / 4. * You run your own scraping code on these platforms. Open this workload in the calculator to see today's prices or change the inputs.

• ScrapingBee is now cheapest at $249. Rendering and premium proxies multiply its credit cost, but the 3 million credits of its Business plan still cover the month.
• ZenRows goes from $99 to $749. The reason is the US requirement: ZenRows only offers country targeting on premium proxies, so every request costs at least 10 credits, and 25 with rendering. ScrapingBee can target a country with its standard proxies.
• Zyte's range is wide, from $200 to $520, because the price depends on the difficulty tier Zyte assigns to your site, which you only learn from Zyte's own estimator.
• Browserless is the most expensive: every rendered page uses browser time, and residential proxy traffic is charged in the same units, 6 per megabyte.

Hard: every page rendered, heavy protection

All pages need a browser, and the site sits behind a commercial anti-bot system.

Assumptions (low / expected / high): success rate 92% / 85% / 75%; pages needing premium proxies 50% / 70% / 90%; website difficulty tier 3 / 4 / 5. * You run your own scraping code on these platforms. Open this workload in the calculator to see today's prices or change the inputs.

• Bright Data is now cheapest, at the same $651 as for easy pages. Its price per successful request includes rendering and proxy selection, and failed requests aren't billed, so difficulty doesn't move it.
• Oxylabs Web Scraper API costs $999, although 500,000 rendered results at its Business rate of $1.00 per 1,000 add up to about $500. Its Starter plan tops out at $598 of budget ($99 plus up to $499 in top-ups), which buys about 460,000 rendered results at $1.30 per 1,000. That is not enough, so the next step is the $999 Business plan.
• ScrapingBee can't take this workload on a self-serve plan. It would need about 9.5 million credits a month, and its largest self-serve plan has 8 million. Above that is enterprise pricing on request.
• Zyte has no high estimate here: at the hardest tier the month costs more than the $2,500 spending limit of its largest self-serve commitment, which again means enterprise pricing.

What actually sets the price

• The billing unit. A flat price per successful request (Bright Data) doesn't care how hard the page is. Credits (ScrapingBee, ZenRows, ScraperAPI) multiply with rendering and proxy type. Traffic (Oxylabs Web Unblocker) grows with page weight. Browser time and compute (Browserless, Apify) grow with every rendered second. Knowing which unit you pay in tells you more than the headline price.
• Plan steps and caps. Top-up limits, credit ceilings and spending limits decide when you have to move up a plan, as the Oxylabs, ScrapingBee and Zyte examples show. A plan that is cheap per page can still be the wrong one if the month doesn't fit it.
• What counts as a success. Most of these products bill only successful requests, but they define success differently. Oxylabs bills target responses with 2xx and 4xx codes, and Zyte bills a 404 page like any other. Product pages list the statuses each provider bills.
• What you can't know in advance. Success rate, the share of pages that need premium proxies, and Zyte's difficulty tier move the estimates most. Measure them on a trial before committing to a plan.

Price your own workload

These are estimates for one workload on one day, not quotes; the provider's checkout is what you will actually pay. The calculator runs the same engine on current prices for your page volume, rendering share, country and page size, and the recommendation page adds limits and features. Agents can do the same through the API and MCP server.

ScrapingBee, Zyte and the other providers here did not sponsor this comparison, and ScrapeSelect earns nothing from any of them. See the disclosure for how that would be handled if it changes.