Four models built the same page.
Could you tell which cost 12× more?
Same prompt, four models, one shot each — every page fully works. We stripped the labels. Click the one you think ran up the biggest bill, then reveal the receipts. (Most people can't.)
You couldn't tell them apart. Neither can your users.
Four pages. A 12× spread in what they cost to build. Quality you can't rank by looking.
Most teams are hard-coded to the expensive end — paying frontier prices for work a far cheaper model nails just as well. On a page like this, that's the whole bill, wasted.
Build a complete, production-quality landing page for "Fairway Social", a neighborhood golf-simulator bar in South Boston. One self-contained HTML file · sticky nav · hero + booking CTA · 4+ amenities · pricing (Off-Peak $35/hr · Prime $55/hr · League $89/mo) · booking form with validation + success state · 3 testimonials · 4-question FAQ · footer · mobile responsive · working dark-mode toggle · tasteful motion. Return ONLY the HTML.
The verdict
Every page fully works. Booking form validates and confirms, dark mode toggles, nothing breaks on a phone. On the functional audit it's a four-way tie — which is exactly why we hid the labels.
So the only thing that separates them is the bill, and the bill ranges 12×. You're looking at four pages you genuinely can't rank by quality, priced from six cents to seventy-seven.
Pick by brand and you're gambling with your budget. Pick by the cheapest that still clears the bar — which is all routing is — and you keep the result while cutting the bill to a fraction. "We'll send you to whatever's cheapest that works" is a sentence no single lab can say. We can.
All four passed the same functional audit — so this is a pure price race. Cost = the delivered page's output tokens × the model's published rate.
| Page | Cost to build | vs cheapest |
|---|---|---|
| Page BCHEAPEST | $0.0646 | 1× |
| Page A | $0.3219 | 5.0× |
| Page D | $0.3799 | 5.9× |
| Page CPRICIEST | $0.7733 | 12.0× |
Four leading models built these — Fable 5, Opus 4.8, GPT-5.5, and one more. We're not telling you which built which, because you couldn't tell either. That's the point.
One shot each
Identical prompt, no cherry-picking. Each page is exactly what the model returned.
Audited by a robot
Every page checked in a real browser: renders clean, correct pricing tiers, a working validated booking form, dark-mode toggle, mobile-responsive at 375px. All four passed.
Real bills
Cost = the delivered page's output tokens × each model's published rate. Same yardstick for all four.
Labels withheld on purpose
We don't map model to page. If quality were the differentiator you'd spot the expensive one. You can't — so it isn't.
Your production traffic has a bill like this too.
We route every call to the cheapest model that clears your quality bar — and prove it nightly. What you spent, what you'd have spent, receipts attached.