llmgauge

price history archive

Price history

An append-only record of what models cost, captured every other day. The value is the start date — this archive only deepens. Track any model below, or watch the floor price of frontier intelligence over time.

cheapest premier model · output / 1M
$0.20
across 166 first-party-priced models · snapshot
cost of intelligence — cheapest premier output price per capture date (archive begins 2026-06-14)
input output

Append-only · every-other-day cadence · JSON: /history. How we measure →

Why the archive compounds

Anyone can publish today's prices. What compounds is a trustworthy record of how they change — captured consistently, sourced, and impossible to quietly rewrite. Every run appends a dated snapshot to an append-only table (additions only, no edits or deletes, enforced in the database), so the longer this runs the more it's worth. Because the cadence is every other day, the series is a sequence of capture dates with gaps by design — not a guaranteed row per calendar day.

Why the floor falls

The headline line tracks the cheapest output price among premier models. It rarely moves because an existing model got cheaper — it moves because a new, capable, cheaper model ships. That's the real shape of "intelligence per dollar": not gradual discounts, but step-changes as the frontier widens. Pair it with measured cost-per-task to see whether a cheaper rate actually means a cheaper job.

Frequently asked questions

How far back does the history go?

The archive is append-only and began on 2026-06-14. It captures a price snapshot every other day, so the series has dated gaps by design — it grows more valuable the longer it runs. Rows can be added but never edited or deleted (enforced at the database level).

What is the 'cost of intelligence' line?

It tracks the cheapest output price among premier (first-party-priced) models on each capture date — a floor for name-brand capability. As cheaper capable models ship, the floor drops; that decline is the story list prices hide.

Related