How should I structure a fair comparison?

Keep one variable: same data, same machine, same tab. Run tasks repeatedly and look at steady-state times rather than the first run, which pays JIT and cache warm-up costs.

Why do my numbers differ between runs?

Browsers share CPU with the OS, other tabs and background work — a few percent of noise is normal. Trust differences that persist across multiple runs, not single victories.

Can it benchmark network calls?

It measures wall-clock duration of whatever you define as a task, including async work — but network variance will dominate. For network comparisons, run many iterations and compare medians.

Benchmark builder — Guide

Easily compare execution time of tasks with this very simple online benchmark builder.

What it is and when to use it

A simple online benchmark builder: define tasks (either a piece of code or a fixed duration), run them, and compare execution times in one table. Built for the classic "which approach is faster" question — with enough structure (labels, iterations, relative bars) that results are comparable rather than vibes.

Times several code snippets side by side to see quickly which implementation is faster.

How to use it

Add the implementations you want to compare
Click run
Read the per-snippet timings and relative ratios

A worked example

Fill in a sample suite name

JSON parse 10k rows

Copy this into the tool, or open Benchmark builder

Things that trip people up

Browser-side timing is affected by machine load — avoid drawing conclusions from differences under ~10%

Questions people ask

How should I structure a fair comparison?: Keep one variable: same data, same machine, same tab. Run tasks repeatedly and look at steady-state times rather than the first run, which pays JIT and cache warm-up costs.
Why do my numbers differ between runs?: Browsers share CPU with the OS, other tabs and background work — a few percent of noise is normal. Trust differences that persist across multiple runs, not single victories.
Can it benchmark network calls?: It measures wall-clock duration of whatever you define as a task, including async work — but network variance will dominate. For network comparisons, run many iterations and compare medians.

Next steps

Open the Benchmark builder tool

Spotted something wrong on this page? Tell us via Contact Us