Pick the one that suits how you like to decide.
Same Veloċi, five ways to get to know it. Switch whenever you like.
Evidence · Automated engineering benchmark · United States
12.4s
on the laptop
Revalidation required
12.5s
on the phone
6
scripted runs
12.4 s on the laptop · 12.5 s on the phone
What you need to judge it
Front door to a finished site on screen
Result12.4 s on the laptop · 12.5 s on the phone
Spread12.0 s to 12.6 s across 5 runs; one outlier of 27.7 s (florist on the phone: the script waited 18.0 s, 15.3 s of it for the builder page to finish loading)
Sample6 scripted runs: 3 business types (cafe, florist, plumber) on a 1366×768 laptop and a 390×844 phone
MethodThe 'typical' scenario: the interview answered, business name, tagline, contact details, one style change, one accent colour, one library photo. 'First site' is the moment the preview loader clears and the site is on screen.
ConditionsChrome (Playwright) on the deployed builder at veloci.build over an internet connection (name lookups kept warm, because the test machine stalls about 12 s on a cold one), typing at 30 ms a key; every write to the backend was answered in the browser and the Stripe hand-off stubbed, so nothing was ordered or charged
When2026-11-08
Which buildbuilder speed1108d, 8 November 2026
What this does not establishScripted wall-clock time is a lower bound on a person: nobody reads, hesitates or decides. It measures how long the product itself takes, its own waits included, not how long a customer takes.
This suite completed 6 of 6 journeys on the deployed build. An earlier suite, in September 2026, reported a phone median far above this one because three runs stalled at the setup wizard; that was a real defect — a first-run tour overlay could open on top of the wizard and swallow every tap — and it is fixed, not excluded.
RevalidationRevalidation required. The builder has changed since this was measured (tested against 1a3a3eddadf1, now 59d306177370).
The number on this page is what was measured then, and is not presented as current.
Underlying evidence
Recorded in the register, not yet published on the site.
Not published yetRecorded in the register, not yet published on the site.
Not published yetRecorded in the register, not yet published on the site.
Citing this
https://veloci.build/evidence/first-site is the address for this figure.
It carries the method and the limitation; a marketing page carries neither. If the number changes, this page changes with it.
Fingerprints are sha-256 (first 12) of veloci-builder.html and of compare-src/build.mjs at the time of the run.
Partner programme
That's the Starter bundle fee. They pay the same, and we never pay for reviews.