Bright Data
brightdata.comScore
TBD
Enterprise proxy + scraping infrastructure. Web Unlocker, Scraping Browser, SERP API.
Best track
Pending
Pricing
Pending
fetchrendersessionextract
Mock data previewTemporarily, this page contains mock data. Real benchmark results are being prepared and will be updated soon; public scores are hidden until then.
Mock previewReal data coming soon
Scraping API benchmarks with visible evidence.
The interface is live, but the benchmark values are placeholders while the first real run is prepared.
Run controlPublic
Current exported benchmark contract.
Runmock preview
Statusscores hidden
CorpusPending
CommitPending
ScoresPending
Current leader
Pending
Real benchmark data is being prepared and will replace these placeholders soon.
Attempts
Pending
Attempt counts are hidden while the site is in mock-data preview.
Coverage
Pending
Provider and track coverage will be published with the real run.
Validator
Pending
Validator metadata will ship with the first real export.
Mock trend preview
Scores hidden · JS Render
How to read it
Success
Share of attempts that returned usable content, not just a transport-level success.
Latency
Median end-to-end response time for successful attempts in that track.
Cost
Estimated provider spend normalized to a standard successful-attempt volume.
Retries
Average attempts needed before a run is marked usable or failed.
Evidence
Per-target logs, timestamps, failure reasons, byte counts, and hashes.
Data feed
Source
Pending
Contract
Pending
Score rows
Pending
Attempts
Pending
Validators
Pending
LLM audit
Pending
Runner writesWebsite readsScores shownEvidence retained
Leaderboard
Provider preview
Scores are intentionally hidden while mock data is on the page. Use this as a preview of the comparison workflow.
| Status | Provider | Score | Open |
|---|---|---|---|
| Preview | TBD | ||
| Preview | TBD | ||
| Preview | TBD | ||
| Preview | TBD | ||
| Preview | TBD | ||
| Preview | TBD | ||
| Preview | TBD | ||
| Preview | TBD | ||
| Preview | TBD | ||
| Preview | TBD |
Scores hiddenTrack cells are placeholders until the first real benchmark run is published.
Browse by job
Benchmark tracks
Scraping APIs are evaluated by the job they are meant to solve, so simple HTML APIs are not judged like hosted browsers or agent workflows.
Track 1
HTML scrape
URL in → HTML/text out. Static pages, no JS, simple anti-bot.
Corpus pending
Track 2
JS Render
Pages requiring a real browser to populate the DOM.
Corpus pending
Track 3
Browser Session
Multi-step CDP flows: click, fill, scroll, persist cookies.
Corpus pending
Track 4
Structured Extract
Field-level accuracy against ground-truth schemas.
Corpus pending
Track 5
Agentic
LLM-controlled browser tasks. Non-deterministic; reported separately.
Corpus pending
Methodology
Corpus and scoring
Same URLs, same parameters, same published run for every provider.
Same target set
Every provider receives the same sampled URLs and task definitions for the published run.
Track-specific adapters
Fetch APIs, render APIs, hosted browsers, extraction APIs, and agent flows are scored separately.
Usable-output scoring
A response only counts when the returned page or fields are usable for the intended scraping task.
Disclosure-first
ScrapeDrive is operated by the same team, appears in the benchmark, and is not excluded when it loses.
| Anti-bot vendor | URLs |
|---|---|
| Akamai-protected | TBD |
| Cloudflare | TBD |
| PerimeterX | TBD |
| DataDome | TBD |
| Unprotected control | TBD |
| Total sampled | TBD |
| Target | Protection | Category | URLs |
|---|---|---|---|
| Amazon · product detail | Akamai | ecom | TBD |
| LinkedIn · company page | PerimeterX | social | TBD |
| Indeed · search results | Cloudflare | jobs | TBD |
| G2 · software listing | DataDome | review | TBD |
| Yelp · business page | PerimeterX | local | TBD |
| Zillow · listing detail | Akamai | realestate | TBD |
| TripAdvisor · hotel detail | Cloudflare | travel | TBD |
| eBay · search results | Akamai | ecom | TBD |
| Static news article (control) | control | control | TBD |
| SPA-only shop (control) | control | control | TBD |
Provider index
Scraping APIs in the benchmark
A browsable directory for comparing provider fit, pricing shape, and supported modes. Scores will appear after the real benchmark run.
Zyte API
zyte.comScore
TBD
Smart proxy with auto-rotation, ban detection, JS rendering. Scrapy creators.
Best track
Pending
Pricing
Pending
fetchrenderextract
Firecrawl
firecrawl.devScore
TBD
URL → clean Markdown for LLMs. Crawl, scrape, extract.
Best track
Pending
Pricing
Pending
fetchrenderextract
ScrapeDriveours
scrapedrive.comScore
TBD
Lightweight scraping API. Built by the team that runs ScrapingEvals.
Best track
Pending
Pricing
Pending
fetchrenderextract
ScrapingBee
scrapingbee.comScore
TBD
Headless browser API with proxy rotation. Aimed at developers.
Best track
Pending
Pricing
Pending
fetchrender
ScraperAPI
scraperapi.comScore
TBD
Rotating proxies + headless browsers. High-volume general scraping.
Best track
Pending
Pricing
Pending
fetchrender
Browserless
browserless.ioScore
TBD
Hosted Chrome over CDP. Best for browser sessions and agentic flows.
Best track
Pending
Pricing
Pending
rendersession
Scrapfly
scrapfly.ioScore
TBD
ASP anti-scraping bypass, screenshots, sessions, extraction rules.
Best track
Pending
Pricing
Pending
fetchrenderextract
Scrape.do
scrape.doScore
TBD
Simple URL-in / HTML-out API. Cheap, broad coverage.
Best track
Pending
Pricing
Pending
fetchrender
Apify
apify.comScore
TBD
Actor marketplace + crawler infrastructure. Wide capability surface.
Best track
Pending
Pricing
Pending
fetchrendersessionextract
FAQ
Common questions
What is ScrapingEvals?
ScrapingEvals is a public benchmark for scraping APIs. It compares providers on real-world scraping jobs instead of only listing features or pricing pages.
Why are there separate tracks?
A raw HTML fetch API, JavaScript renderer, persistent browser session, structured extractor, and agent-controlled browser solve different jobs. One overall score would hide those differences.
How should a beginner read the leaderboard?
Start with the track that matches your job. HTML scrape is for static pages, JS Render is for browser-built pages, Browser Session is for multi-step workflows, Structured Extract is for field accuracy, and Agentic is for LLM-driven browsing.
Can the results be reproduced?
The site links run ids, commit hashes, adapter source, target categories, and evidence logs so another engineer can inspect how the numbers were produced.
Is ScrapeDrive favored?
No. ScrapeDrive is marked because the ScrapingEvals team operates it. The same tables, evidence view, tracks, and scoring language apply to every provider.
Reproducibility
Published artifacts
Run id
mock-2026-05-17
Commit
Methodology
scoring-v1
Adapters
10 providers
Tasks
Pending
Artifacts
Pending