PRICING / START WITH THE WORKLOAD

Pricing that starts with a benchmark, not a guess.

The final Dvar rate card is still being shaped with early enterprise partners. The structure is clear now: start with an eval, pay for the traffic you run, and only commit to capacity when the workload proves it.

Book a workload review See the pricing shape

PROVEEvalsMake quality visible.

ROUTEDvar-DynamicPay for traffic that clears.

OWNDvar ModelsInvest where quality repeats.

ILLUSTRATIVE STRUCTURE

Simple enough to compare.
Flexible enough to fit.

These are placeholders for the first public rate card. We will publish concrete model and compute rates after validating demand, regional costs, and support requirements with design partners. Support is part of the proposal, not a surprise line after go-live.

PRODUCTPLANHOW IT BILLSWHAT IS INCLUDED

Dvar EvalsWorkload reviewScoped engagement

A quality, latency, and unit-cost baseline built with your team from representative traffic.

Dvar-DynamicProductionPer 1M tokens

Usage-based routing to the cheapest model that clears your eval, with serving and cost visibility.

Dvar ModelsCo-buildScoped engagement

Evaluation, data curation, training, deployment, and an operated improvement loop when volume earns it.

PILOT

Start with a workflow.

For a team proving a use case with governed workspace access or a shared inference endpoint.

Focused onboarding and model shortlist
Usage and quality review
Clear path to production

Discuss a pilot

PRODUCTION

Scale the path that works.

For a product team that needs predictable capacity, policy controls, and a production operating rhythm.

Shared or dedicated inference
Metering and deployment support
Volume pricing as usage grows

Discuss production

ENTERPRISE

Co-build the advantage.

For a strategic workflow where data, evaluation, and specialized models create a durable quality wedge.

Fine-tuning and evaluation loop
Dedicated support and governance
Custom commercial terms

Talk to Dvar

THE HONEST ANSWER

We will not hide the operating cost.

Model API price is only one line on the bill. A useful decision includes tokens, latency, GPU occupancy, support, data boundaries, and the cost of a bad answer. That is what the workload review is for, and our team stays accountable for the path after the first deployment.