Scope & pricing

Know what you’re buying.
Before you commit.

Separate the deployment decision, the implementation, and the cost of running it. All prices are USD. Infrastructure and third-party fees are additional.

Establish the scope

Deployment readiness sprint

$2,500fixed fee

A clear technical scope and implementation decision.

  • Infrastructure boundary and dependency map
  • Existing compute review or comparison of up to two providers
  • Model compatibility, workload test plan, and operating-cost range
  • Named owners and a go / no-go implementation scope

One workload and one target environment. Assessment and planning; no installation included.

Discuss your deployment

Readiness is optional when your environment, implementation scope, and acceptance criteria can already be established. If commissioned, the sprint is a separate fee, with no automatic credit toward implementation. Sprint plus workflow implementation: $15,000–$22,500 before infrastructure, third-party fees, and taxes. Scope and acceptance criteria are agreed before kickoff.

Already know the scope?

Deployment-only starter

For a qualified environment where your team owns the application integration.

$7,500one-time implementation

Read the narrower scope

What the $7,500 starter includes

One approved model. One existing environment. One synchronous inference interface. It provides:

  • Model and dependency inventory
  • Authenticated synchronous inference interface
  • One documented HTTP client example
  • Agreed access and workload checks
  • Versioned deployment files, tests, and operator runbook

Qualification before the fixed fee applies

  • An approved model license and supportable runtime
  • An approved Linux GPU environment and working drivers before implementation; sourcing is available separately
  • Approved access and a named infrastructure operator
  • A representative non-sensitive workload and written acceptance criteria

A separate readiness sprint is not required when these prerequisites and the scope can already be established. If a sprint is commissioned, its fee is additional.

Separate scope

  • Cloud, hardware, third-party fees, and taxes
  • Application integration beyond the HTTP client example
  • Model release security scans, signatures, and registry promotion
  • Durable queues, persistent results, and data migration
  • High availability, multi-node serving, and 24/7 operations
  • Post-training, clinical use, patient data, and compliance certification

Additional application paths, preprocessing, and model-specific external services are quoted separately.

Download the deployment-only scope worksheet →

Pay against a written delivery scope.

Starter payment: $3,750 at kickoff and $3,750 after acceptance. Sprint and workflow billing milestones are agreed in the written scope.

30-day correction commitment: in-scope implementation defects reported within 30 days after acceptance are corrected without another implementation fee. This covers the starter’s endpoint and HTTP example, or the workflow implementation’s agreed application connection. Review and correction periods, blocked prerequisites, cancellation, and refund terms are recorded before kickoff.

The commitment covers implementation defects, with audit outcomes, model accuracy, and ongoing availability outside its scope. Review the acceptance criteria.

Compute and release security.

Compute sourcing can be part of the $2,500 readiness sprint: one target environment and up to two provider comparisons. Provider capacity, hardware, colocation, and infrastructure charges are paid separately.

Model release security is quoted per model and runtime. Artifact scans, runtime CVE evidence, signatures, and registry promotion are not included in either implementation package by default.

Keep releases current.

$1,000–$2,500 / month, optional. Each plan specifies 4–10 business hours per month for agreed updates, release tests, and maintenance tasks. Price depends on the hours and tasks selected; response expectations are written into the plan.

Your operator owns the service outside that allowance. Maintenance excludes unlimited incidents and 24/7 on-call coverage. No Tenantplane management subscription is required.

Illustrative scenarios · September 9, 2026

What does the first year cost?

Start with a worked example, then replace the assumptions. Hosted GPU rates below are comparisons—not quotes for your cloud, or proof that your model fits. All calculations stay in your browser.

Hosted L4 comparison: $0.80 × 730 billed hours = $584/month. Assumptions: $100 storage/network and 4 operator hours × $125 = $500/month. These storage and labor amounts are planning assumptions.

$29,208.00 estimated first year · $1,184.00 recurring per month. Example: $12,500 workflow implementation + $2,500 readiness sprint.

Formula: implementation + readiness + other one-time costs + 12 × (compute + storage/networking + operations). A 730-hour month is a planning convention. Taxes, redundant capacity, staging, extraordinary transfers, and licenses are excluded unless entered. Changed inputs invalidate the download until recalculated.

GPU source: Hugging Face hosted AWS endpoint pricing, checked 2026-09-09. Billed hours and actual capacity must be verified for your workload. Presets replace monthly inputs only; implementation, readiness, and other one-time costs remain as entered.

See it on your cluster

Let your teams share GPUs
without the help desk.

Book a 30-minute walkthrough of your GPUs, teams and workloads.

Book a demo