Calculator controls become available when the local script loads. Until then, the readout shows only the starting example.
Price references & assumptions
Reference inputs checked 29 October 2026. Standard paid, uncached text processing, no discounts. Prices are dated references, not a quote, live inventory or capacity advice.
Runpod pricing ↗H100 PCIe: $2.89 per GPU-hour. Cloud tier, region and availability require confirmation.
CoreWeave pricing ↗North America on-demand: 8 × H100 at $49.24 per node-hour; 8 × B200 at $68.80 per node-hour. Each unit is a whole eight-GPU node.
Model usage uses 30 days. GPU billing uses the independently selected hours per capacity unit. Total equals unrounded model and GPU usage plus a fixed 35% management fee.
Setup, integration, storage, network, licences, resilience and taxes are excluded or scoped separately. Media, tools, retries, cache effects, regional variations and other charges are not modeled. Charged reasoning belongs in the billable output input.
GPU rental does not grant proprietary model access. Providers and hardware are not performance-equivalent. External model calls involve external processing; private hosting and residency require separate design. Final scope, supplier availability and commercial terms require confirmation.