Public benchmark · Proof, not promises

A million scrapes.
In public. Live.

Every week, any visitor can trigger a run that scrapes 1,000,000 real public posts in about 20 minutes — roughly 72 million a day at that pace. Watch the success rate, throughput and latency move live. No other scraping API lets you do this.

03days 14hrs 22min 08sec Public record online
Fresh corpus every run Every result downloadable Published forever
The live console

What you see
while it runs.

Public benchmark waiting for a run 00:00:00

Run this week's benchmark

Building a fresh corpus 0 / 0 profiles

Crawling the verified profile corpus for recent public posts and assembling 250,000 active URLs plus 25,000 reserves before the measured waves begin.

Completed
0 / 1,000,000
0.0% of the run
Success rate
0 ok · 0 failed
Throughput
/s
Latency
p95
Scrape progress0 / 1,000,000 posts
By platformlive
Live resultsstreaming
Run complete

Benchmark complete

Finalizing results.

Success rate
Completion /s
Total scrapes
Avg latency
Total time
What we scraped
See the dataset

Benchmark stopped

The run did not complete. Its partial results remain on the public record.

This console is connected to the public benchmark controller.

Why this exists · How it works

Don't trust us.
Test us.

We run millions of scrapes a day for production teams — a public run clears its 1,000,000 in about 20 minutes, roughly 72 million a day at that pace. Before you trust a vendor with that volume, you should watch it clear a real load and check the numbers yourself — so any visitor can put the network through a 1,000,000-scrape run, once every 7 days.

  1. 1
    Any visitor triggers it

    Make a free account ($10 credit, no card) and fire the run yourself — no sales call, no demo. One run at a time, one every 7 days, so it stays fair.

    Run the benchmark 1 run · every 7 days
  2. 2
    Choose the dataset

    You decide which million links we scrape:

    Our standing dataset

    A validated library — 250,000 active URLs plus 25,000 reserves. Four waves, 1,000,000 scrapes, starts instantly. Download and inspect it first.

    Build a fresh corpus

    Our profile scrapers crawl verified profiles for fresh posts, build a brand-new 275,000-link set, and run four waves — 1,000,000 scrapes.

  3. 3
    We scrape all 1,000,000 — live

    Every request takes the same production path as a paying customer's — residential network, real-visitor traffic, per-platform parsers. No special harness. You watch the console tick up live: throughput, success rate, latency.

    Live812 /s
    620,400 / 1,000,00099.4%
  4. 4
    Cross-check every result

    Every row is published with a link to its live post — open any of them and check our number against the source. Download the full set as JSON or CSV.

    @natgeo1.28M views
Built for buyers doing diligence. Weighing Refetcher for millions of scrapes a month? This is your load test — public, repeatable, and yours to run before you pay a cent.
Talk to us about volume
The dataset

Every row is real,
and you can check it.

Both halves of the run, yours to inspect: the links we point at and the results we got back. Pull a random 10,000-row sample to spot-check, or download the whole set — either half, as JSONL or CSV.

A fresh random draw — links and their results, side by side. Shuffle to redraw.
Input

The links we point at

The post URLs this run was handed — the dataset we ran the test on, before any request goes out.

Linkset · sample 275,000 active + reserve URLs
PlatformTargetSource URL
The linkset loads here the moment a run is armed.
Output

The results we hand back

What each link returned — the metric, latency and status. Open any row to check it against the live post.

Results · sample 0 streamed rows
PlatformTargetCapturedScraped at (UTC)LatencyStatus
Rows stream in here live as the run scrapes — each one links to its real post.
The public record

Every run,
on the record. Forever.

Anyone can copy a landing page overnight. Nobody can fake a year of weekly runs. Every run gets a permanent page — including the weeks it dipped, incident note attached.

View all runs on the record
No staging, no spin

What a run
refuses to do.

A number is only as honest as the shortcuts it couldn't take. Here's what a public run is built so it can't do.

01

No warm-up

A visitor fires it at a random hour, into whatever live load the network is already carrying. There's no quiet moment picked to make the numbers look better.

02

No fast lane

Every request takes the same production path a paying customer's does — the same residential network, the same parsers. There's no benchmark-only shortcut.

03

No hand-picked targets

The corpus is assembled from a rotating profile list, not a curated set of posts we already know will score well. We don't get to choose the easy ones.

04

No smoothing the numbers

Each row carries the raw captured value and its week-over-week delta — no rounding up, no "up to" figures, no averages hiding a bad platform inside a good total.

Straight answers

The benchmark, in plain English.

What is the Refetcher public benchmark? +

Every week we scrape 1,000,000 public posts in the open. Anyone can watch the run live and audit the results after — the only scraping API with a benchmark you can watch happen.

How fast is a run, and how much can Refetcher scrape in a day? +

Each public run scrapes 1,000,000 posts in about 20 minutes — roughly 72 million a day at that pace. It runs on the same production path that serves paying customers, who run millions of scrapes a day.

Does the benchmark slow down paying customers? +

No. Benchmark traffic is deprioritized below paying customers, so a run can't degrade a real customer's success rate or latency — even though it runs at a random, real-world hour.

Who can trigger a run? +

Anyone with a free account — $10 credit, no card. One run at a time, one per week; everyone else watches the same live console.

Don't take our word for it.

Make a free account, trigger a run, and watch a million scrapes clear in real time. The proof is the product.

Create a free account