Benchmarks

Same sites. Measured openly. Not always the same task.

Our rows are full structured extraction through this API. Competitor $/1k figures that say “base scrape only” are HTML or markdown fetch — they are not like-for-like with JSON extraction.

Honesty first: numbers below are from a live run on 25 Aug 2026 (evening batch) against the public URLs listed in the methodology. Cost uses the 1,000-page pack list price ($16 ÷ 1,000 pages = $0.016/page). Re-run any row yourself with the same instruction — we will never publish a number you can't check.

Live results · 25 Aug 2026

What we measured

Seven public site categories. One extraction task each. End-to-end until a terminal job status.

Site class	Task	Rows correct	Time to dataset	Cost to dataset	Run id
E-commerce	Products w/ price, rating, stock	40/40	68s · 2 pages	$0.032	`job_01a03943e545d7aea3e0778646e758bf`
Encyclopedic	Structured facts from an article	1/1 record	61s · 1 page	$0.016	`job_01a03944ebca7c71973f69e35ca6e028`
Code hosting	Repo metadata + latest release	1/1 record	40s · 1 page	$0.016	`job_01a03945d93d8c471de9a9659709037c`
Forums	Threads w/ author, points, replies	30/30 titles · points/author sparse	35s · 1 page	$0.016	`job_01a03946755507d7d9a46b35a7f1b505`
Storefront platforms	Catalog across pagination	60/60	118s · 3 pages	$0.048	`job_01a03946ff99a5bb7a75ac31d9009800`
News	Posts w/ title, date, link	8/8	111s · 2 pages	$0.032	`job_01a03948cd59a50cf2522fd8f7631739`
Directories	Listings w/ contact fields	12/12	64s · 1 page	$0.016	`job_01a0394a80938a3493c21bf32b7b5665`

Method on these runs was render except encyclopedic/code (llm) and directories (api). Scoring: required fields present on returned rows vs. the site's visible listing (hand-checked sample + expected row count where the page size is known). Time is wall-clock from submit to terminal poll. Cost = pages_used × 1,000-page pack rate ($16 ÷ 1,000). HN returned every story title/link but left points and author empty on this run — reported as-is. HN robots.txt carries Crawl-delay: 30; it did not add wall time on this 1-page run (delay applies between pages).

Known open (forums / HN): Hacker News exposes public APIs (hn.algolia.com/api/v1/search?tags=front_page, Firebase topstories) that already include points and author. This run did not use those endpoints, so points and author came back empty.

Methodology

URLs and instructions you can copy.

Same API the product ships. No special benchmark lane.

Class	Endpoint	URL	Instruction / question	max_pages
E-commerce	`/v1/extract`	`https://books.toscrape.com/`	every book with title, price, star rating and availability	2
Encyclopedic	`/v1/extract/page`	`https://en.wikipedia.org/wiki/Web_scraping`	definition, main techniques, common uses, and legal or ethical concerns as separate fields	1
Code hosting	`/v1/extract/page`	`https://github.com/psf/requests`	repository name, description, primary language, stars count, forks, license, and latest release tag if shown	1
Forums	`/v1/extract`	`https://news.ycombinator.com/`	every story with title, points, comment count, author and link	1
Storefront platforms	`/v1/extract`	`https://books.toscrape.com/catalogue/page-1.html`	every book with title, price, rating and availability across catalogue pages	3
News	`/v1/extract`	`https://blog.python.org/`	every post with title, date, and link	2
Directories	`/v1/extract`	`https://www.mdd.gov.hk/en/mdacs/search-database/list-distributors/index.html`	every licensed distributor with English name, Chinese name, phone and expiry date	2

Verifiable today

The cost math you can check yourself.

Published list prices. Rows marked “base scrape only” or “no AI extraction” are fetch-only — they are not the same job as Plainscrape’s structured JSON. Per 1,000 pages:

Provider & configuration	Model	Effective $/1k pages
Firecrawl Standard — base scrape only, no extraction	1 credit/page	$0.83
Bright Data Unlocker — PAYG, no AI extraction	Per successful request	$1.50
ScrapingBee Freelance — JS + AI query (10 credits)	Credit multipliers	$1.96
Firecrawl Standard — + JSON extraction (5 credits)	Credit multipliers	$4.15
ScrapingBee Freelance — premium proxy + JS + AI (30 credits)	Credit multipliers	$5.88
ScrapingBee Freelance — stealth + AI (80 credits)	Credit multipliers	$15.68
Plainscrape — 1,000-page pack	Flat per page, extraction included	$16.00
Firecrawl Hobby — + JSON extraction (5 credits)	Credit multipliers	$16.00

On published per-page rates, Firecrawl Standard (base scrape) and ScrapeGraphAI Growth are cheaper than our listed $16/1k pack. We tie Firecrawl Hobby on JSON extraction at the same $16/1k when you use the plan's full monthly capacity. Below roughly 1,000 pages monthly, rivals' subscription floors hurt them more than our flat per-page rate; at 2,000 pages a month ScrapeGraphAI Starter ($20/month) is cheaper, at $10.00 per 1,000 extractions. Above that we lose on published rival rates alone. Checkout sells the 1,000-page pack; volume above 1,000 pages is quoted at enterprise@plainscrape.com with no listed per-page rate. Extraction, rendering and proxying are in the page price, and unused pack pages stay valid 12 months. Rates verified 30 Aug 2026 against each vendor's public pricing page; check theirs before you commit to either of us. Spot an error? Email hello@plainscrape.com.

Re-run any row in the playground.

Try it in the playground

Documented methodology · live receipts above · corrections welcome