PRICING / DEPLOYMENT ECONOMICS

Pay for the operating path
your workload actually needs.

SPARK pricing follows the resource that drives the work: tokens for elastic access, accelerator time for dedicated capacity, and a designed service plan for private grids and model programs.

Compare paths
01 / ELASTIC API · PRIVATE PREVIEW

Usage based

From $0.60 / 1M input tokens
From $2.40 / 1M output tokens

Introductory SPARK model list rate for approved evaluation accounts.

  • Scoped API credential after review
  • Model-specific rates disclosed before use
  • Standard request telemetry
  • No idle-capacity allocation
Request private access →
03 / PRIVATE GRID

Designed program

From $12,000 / month

Starting engineering and operations plan; compute, network, region, and third-party charges are quoted separately.

  • Architecture and data-flow review
  • Model lifecycle and release plan
  • Contract-specific operations scope
  • Customer-approved capacity proposal
Design a private grid →
WHAT SHAPES THE ESTIMATE

Workload physics
before a price sheet.

MODELSize, precision, modalities, context, and adaptation method

TRAFFICPrompt/output length, concurrency, burst shape, and batchability

TARGETQuality threshold, latency percentiles, availability, and recovery

BOUNDARYRegion, isolation, networking, retention, and access policy

LIFECYCLEEvaluation, training, release frequency, and operations support

COMMERCIAL NOTE

Prices are USD introductory list rates before taxes. They are not a promise that a particular model, accelerator, region, or capacity is immediately available. Provider terms, hardware choice, supplier pricing, support, and committed capacity affect the signed quote.