PRICING

Bespoke doesn’t mean open-ended. Every engagement is scoped and priced up front — hardware, model, and support — so procurement gets a number, not a surprise.

ENGAGEMENTS

Three ways to work with Demirari AI.

Hardware only

Custom machine, DemirOS, burn-in, deploy.

From$48,500one-time
  • Spec design session
  • Built-to-order machine
  • Validation & 72-hour burn-in
  • On-site deployment
  • 1 year Essential support
Request a quote
Most teams start here

Hardware + Model

The full bespoke system: machine + Modelsmith model, co-designed.

From$96,000one-time
  • Everything in Hardware only
  • Fine-tune / distill / quantize / RAG
  • Private weights — you hold them
  • Independent eval report
  • 1 year Priority support
Request a quote

Model only

Modelsmith engineering on hardware you already own.

From$38,000one-time
  • Data curation pipeline
  • Tuning & quantization
  • Held-out evaluation
  • Deploy assistance on your iron
  • 1 year Essential support
Request a quote

Every quote is scoped to your exact build — these are real starting points, not teaser rates.

SUPPORT

Pick the SLA your risk model demands.

Essential

$1,200/mo

  • Response timeNext business day
  • MonitoringBusiness-hours health checks
  • Update cadenceQuarterly DemirOS updates
  • On-site sparesDepot exchange
  • Retraining cadenceAnnual model review

Priority

$3,600/mo

  • Response time4 hours, 12×5
  • Monitoring24/7 monitoring & alerting
  • Update cadenceMonthly updates + security patches
  • On-site sparesNext-day on-site spares
  • Retraining cadenceSemi-annual retraining

Mission-Critical

$7,500/mo

  • Response time1 hour, 24×7
  • Monitoring24/7 monitoring + named engineer
  • Update cadenceContinuous, staged rollouts
  • On-site sparesStocked spare kit, same-day
  • Retraining cadenceQuarterly retraining + drift watch

Year one of Essential support is included with every build. Upgrades are prorated, never penalized.

THE MATH

The cloud meter never stops. Yours does.

At sustained volume, owning a tuned on-prem system typically breaks even against cloud API billing well within the hardware’s life — and you keep the asset.

80M req
5M500M
3 years
1 yr5 yrs
Est. 3-yr cloud
$273,600
Demirari AI
$183,200
Break-even
~22 months
$78k$156k$234k$312kMO 0YR 1YR 2YR 3BREAK-EVEN ~22 MO
Cloud API (cumulative) Demirari AI (up-front, then flat)

Illustrative: blended cloud API rate $95/M requests · typical bespoke build $140k · support $14k/yr. Your quote will be exact.

QUESTIONS

Asked before every signature.

Yes. After the design consultation we scope your build to the component and the model to the milestone, then put a single number on it. If the scope doesn’t change, the price doesn’t change — that’s the entire point.

Request a quote
Open the Configurator