Two workloads. Same model, best hardware for each. 40x apart.

Adaptive placement infrastructure for AI inference

To measure what compute produces, and move every workload to where it produces most.

Read the paper → pip install berth-placement

Two services · identical weights · each already on its cheapest feasible placement

Service A
$1.06
Service B
$41.86

Cost per million served tokens. The difference is workload shape, not hardware. Neither team could see it, because no metric they had was denominated in delivered work.

1.0

AI is the first trillion-dollar infrastructure that cannot yet state, in a standard unit, what its output costs.

That spending buys hours. What anyone actually needs is work delivered before a deadline. No standard unit connects the two, so nobody can say what a workload costs, compare two places to run it, or prove a change made anything better.

Every efficiency gain in AI today is unmeasurable. It is why capital cannot tell a good decision from a lucky one, and why two teams running the same model, each on the cheapest hardware that meets their target, can be 40x apart on what a delivered token costs them.

We built the missing meter, and the system that acts on what it reads.

2.0

The unit

The served tokenSVT

One output token delivered inside its service-level objective and above its quality floor.

price notation  $/MSVT  ·  dollars per million served tokens

A token that arrives after its deadline is not a cheap token. It is not a token.

Every existing measure drops something.

  • $ / GPU-hourdrops the output
  • $ / raw tokendrops the deadline and the quality
  • benchmarksdrop the price
  • uptime SLAsdrop everything but the deadline

The served token is the first measure of AI compute that carries all of it at once: the work, the deadline, the quality, and the invoice.

The specification is published, and anyone may use it, including the people who compete with us. A unit nobody else may use is not a unit.
3.0

Adaptive placement

Deciding which chip, which provider, and at which price a model should run under a stated p99, then moving it as prices, capacity, and traffic shift.

The problem: that answer changes. A model version ships. A price moves. Traffic shifts. Most teams decide once and never check.

What makes it adaptive: it never stops deciding. Prices move, models ship, capacity changes. So does the right placement.

4.0

The infrastructure

Three pieces. The unit is measured, the measurement is checked, and the decision is acted on.

berth

Predicts what a workload costs and how fast it runs on a given accelerator, before you rent it.

sounding

Measures the real thing and checks the prediction.

pilot

Watches for change, adapts the placement, and proves what the change saved.

$ pip install berth-placement

All open source. All reproducible from published traces.

5.0

It is arithmetic

The model is a page of physics: bytes moved, work done, time under load. No learned components and no training data, so every number is auditable by hand.

That matters because an answer is only worth having if you can check it. Every prediction ships with the term that produced it and a label saying whether that term rests on a measurement or a specification sheet.

Where the physics stops describing the machine, we will learn the difference. Knowing exactly where it stops is what makes that possible, and mapping that boundary is the research.

6.0

Measured, and published either way

An untuned model, fitted on one accelerator, predicts hardware it has never seen to within ten percent. The bar was set and published before the first measurement was taken.

Every trace is downloadable. Every number is reproducible. When the instrument itself is wrong, we publish that too, with the mechanism and the test that catches it if it returns.

A measurement tool that has never been caught lying has not been used hard enough.
7.0

Working together

The instrument is free and stays free. What we sell is holding a placement correct over time, and proving it.

  • Placement assurance per workload class, per year

    We hold a class against its service level, re-measure when the corpus moves, and issue a conforming receipt each quarter. A class is one model, one serving configuration, one service level.

  • Adaptive placement a share of measured savings

    Above a threshold, we are paid on the difference we prove under the published Holdout Protocol. Below it the holdout costs more than the arrangement returns, assurance is the right instrument, and we will say so.

  • Corpus access per seat, per year

    The measurement record, for allocators, lenders and operators underwriting compute. Including series nobody else publishes.

No provider ever pays. Not for routing, not for ranking, not for placement in the Index. Buy-side only, and every data sale is disclosed.
8.0

Working with us

  • If you serve your own modelsUse berth and tell us where its numbers are wrong. Bring traces.
  • If you operate hardware or a cloudContribute measurement access under our published governance.
  • If you work on performance modelling, systems learning, or market designThis is a genuine open problem with a deterministic verifier and a live environment attached.