scrapingbench

Scrapfly for AI agents: benchmark results

Scrapfly (scraping infrastructure) was tested on scrape and crawl, not on search and map. It got the same tasks, the same model (Claude Haiku 5.5), the same prompt and the same limits as every other provider, and a script checked every answer.

Scrapfly website

Scrape 60 tasks · Read one page and pull out exact fields: prices, versions, install commands.

Tasks done
88%
53 of 60
95% interval
80%–90%
Cost per finished task
0.61¢
Median time
5.0 s

5th of 8 on tasks done. Scrapfly is behind Context.dev (92%), but the gap of 3 points is too small to rule out chance.

Raw output, no model: the API's own response held every answer on 43 of 55 applicable tasks (78%), at 0.02¢ of API spend per finished task.

Tuned settings, frozen before the run: 55 of 60 (92%), against 88% with defaults.

Scrape results, Scrapfly highlighted
ProviderTasks done95% intervalCost per finished taskMedian time
Context.dev55 of 60 (92%)90%–100%0.72¢3.7 s
Linkup54 of 60 (90%)90%–90%0.60¢3.4 s
Firecrawl54 of 60 (90%)90%–90%1.22¢3.3 s
TinyFish54 of 60 (90%)80%–92%0.43¢3.5 s
Scrapfly53 of 60 (88%)80%–90%0.61¢5.0 s
Jina52 of 60 (87%)60%–92%0.57¢3.8 s
Browserbase49 of 60 (82%)78%–100%1.12¢3.3 s
Spider12 of 60 (20%)4%–100%2.48¢4.6 s

All scrape results and charts

Crawl 10 tasks · Read a whole section of a site and return every record in it.

Tasks done
70%
7 of 10
95% interval
0%–95%
Cost per finished task
2.88¢
Median time
14 s

1st of 3 on tasks done. Scrapfly finished the most crawl tasks, but its lead over Context.dev (60%) is too small to rule out chance.

Crawl results, Scrapfly highlighted
ProviderTasks done95% intervalCost per finished taskMedian time
Scrapfly7 of 10 (70%)0%–95%2.88¢14 s
Context.dev6 of 10 (60%)0%–82%7.58¢18 s
Firecrawl5 of 10 (50%)0%–68%14.2¢60 s

All crawl results and charts

Read these numbers with

  • Scrapfly's tuned scrape preset finished 55 of 60, 2 more than its default, but that change is too small to rule out chance.

Scrapfly compared

Other providers