User-supplied performance only

GPU inference cost calculator

Estimate GPU cost per hour, monthly spend and cost per one million tokens. Choose an official cloud rate, then enter throughput measured on your own model, serving stack and workload. This tool contains no benchmark assumptions.

  • Only rates marked live, recent or verified when the catalog was built
  • No default tokens-per-second claim
  • Every selectable rate links to official evidence

Inference scenario

Use your measured workload inputs

Throughput means output tokens per second for one GPU while inference is active. Utilization is the share of each billed hour spent producing tokens. Monthly hours are billable GPU-hours, capped at 744 for this planning view.

Selected official rateChoose a rate from the table belowNo offer is preselected.
Select or change rate
Official cost / GPU-hourSelect an offer
Monthly GPU costAdd billable hours
Cost / 1M output tokensAdd throughput and utilization
Estimated monthly tokensComplete the scenario

Select an offer and enter all three workload inputs. Empty fields are intentional: SaaS Sentinel does not invent inference performance.

JSON-driven rate selector

Choose an official GPU rate from this snapshot

These records were comparable, marked live, recent or verified, expressed in USD per GPU-hour, and linked to an official HTTPS source when the catalog was generated. To keep the calculator fast, the selector retains the lowest published row for each provider, GPU, market and expiry cohort, then keeps only the cheapest still-current cohort at runtime; all 870 regional rows remain in the global comparator and downloads.

61 candidate rates in this view
Official GPU rates marked eligible when this inference catalog was generated
Use rateProvider / GPURegion / marketUSD / GPU-hourFreshnessEvidence
VultrA16 · Ampere 16GBblron demand · 2 GPU/node$0.4710recentObserved 2026-09-09T22:35:58Z · valid through 2026-09-10T04:35:58ZOfficial source ↗
PaperspaceP4000 · Pascal 8GBams1on demand · 1 GPU/node$0.5100recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
Akamai Cloud ComputingRTX 4000 Ada · Ada 20GBap-northeaston demand · 1 GPU/node$0.5200recentObserved 2026-09-09T22:35:26Z · valid through 2026-09-10T04:35:26ZOfficial source ↗
PaperspaceRTX 4000 · Turing 8GBams1on demand · 1 GPU/node$0.5600recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
Microsoft AzureA100 · PCIe 80GBeastusspot · 1 GPU/node$0.6788recentObserved 2026-09-09T22:35:25Z · valid through 2026-09-10T04:35:25ZOfficial source ↗
PaperspaceA4000 · Ampere 16GBams1on demand · 1 GPU/node$0.7600recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
PaperspaceP5000 · Pascal 16GBams1on demand · 1 GPU/node$0.7800recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
PaperspaceRTX 5000 · Turing 16GBny2on demand · 1 GPU/node$0.8200recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
OVHcloudTesla V100S · 32GBus-east-va-1on demand · 1 GPU/node$0.8800recentObserved 2026-09-09T22:35:28Z · valid through 2026-09-10T04:35:28ZOfficial source ↗
ScalewayL4 · 24GBfr-par-1on demand · 1 GPU/node$0.9176Indicative SaaS Sentinel calculation. Native price €0.7875/node-hour ( €0.7875/GPU-hour). USD uses 1.1652 USD/EUR from the ECB reference dated 2026-09-09. ECB reference rate ↗. ECB reference data are available free from the ECB. Not an invoice or transaction rate.recentObserved 2026-09-09T22:35:30Z · valid through 2026-09-10T04:35:30ZOfficial source ↗
Nebius AI CloudRTX PRO 6000 · Blackwell 96GBus-central1spot · 1 GPU/node$0.9500recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
OVHcloudL4 · 24GBus-east-va-1on demand · 2 GPU/node$1.0000recentObserved 2026-09-09T22:35:28Z · valid through 2026-09-10T04:35:28ZOfficial source ↗
DataCrunch (now Verda)RTX 6000 Ada · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$1.0400recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
PaperspaceP6000 · Pascal 24GBams1on demand · 1 GPU/node$1.1000recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
Microsoft AzureH100 · NVL 94GBeastusspot · 2 GPU/node$1.2899recentObserved 2026-09-09T22:35:25Z · valid through 2026-09-10T04:35:25ZOfficial source ↗
CivoL40S · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$1.2900recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
PaperspaceA5000 · Ampere 24GBca1on demand · 1 GPU/node$1.3800recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
ScalewayP100 · 16GBfr-par-1on demand · 1 GPU/node$1.4227Indicative SaaS Sentinel calculation. Native price €1.2210/node-hour ( €1.2210/GPU-hour). USD uses 1.1652 USD/EUR from the ECB reference dated 2026-09-09. ECB reference rate ↗. ECB reference data are available free from the ECB. Not an invoice or transaction rate.recentObserved 2026-09-09T22:35:30Z · valid through 2026-09-10T04:35:30ZOfficial source ↗
HyperstackA100 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$1.4800recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
Akamai Cloud ComputingRTX 6000 · Turing 24GBap-northeaston demand · 1 GPU/node$1.5000recentObserved 2026-09-09T22:35:26Z · valid through 2026-09-10T04:35:26ZOfficial source ↗
Crusoe CloudL40S · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$1.5000recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
DigitalOceanL40S · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$1.5700recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
DigitalOceanRTX 6000 Ada · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$1.5700recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
VultrL40S · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$1.6700recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T04:35:19ZOfficial source ↗
ScalewayL40S · 48GBpl-waw-2on demand · 1 GPU/node$1.7127Indicative SaaS Sentinel calculation. Native price €1.4699/node-hour ( €1.4699/GPU-hour). USD uses 1.1652 USD/EUR from the ECB reference dated 2026-09-09. ECB reference rate ↗. ECB reference data are available free from the ECB. Not an invoice or transaction rate.recentObserved 2026-09-09T22:35:30Z · valid through 2026-09-10T04:35:30ZOfficial source ↗
CivoA100 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$1.7900recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
Nebius AI CloudRTX PRO 6000 · Blackwell 96GBus-central1on demand · 1 GPU/node$1.8000recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
OVHcloudL40S · 48GBus-east-va-1on demand · 2 GPU/node$1.8000recentObserved 2026-09-09T22:35:28Z · valid through 2026-09-10T04:35:28ZOfficial source ↗
PaperspaceA6000 · Ampere 48GBca1on demand · 1 GPU/node$1.8900recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
Nebius AI CloudH100 · NVLink 80GBeu-north1spot · 1 GPU/node$2.1500recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
Crusoe CloudA100 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$2.3000recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
PaperspaceV100 · 32GBny2on demand · 1 GPU/node$2.3000recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
LambdaA100 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$2.3900recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
Nebius AI CloudH200 · NVLink 141GBeu-north1spot · 1 GPU/node$2.4500recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
HyperstackH100 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$2.8500recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
CivoH100 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$2.9900recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
PaperspaceA100 · 40GBny2on demand · 1 GPU/node$3.0900recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
PaperspaceA100 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$3.1300recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
ScalewayH100-PCIe · 80GBfr-par-2on demand · 1 GPU/node$3.3400Indicative SaaS Sentinel calculation. Native price €2.8665/node-hour ( €2.8665/GPU-hour). USD uses 1.1652 USD/EUR from the ECB reference dated 2026-09-09. ECB reference rate ↗. ECB reference data are available free from the ECB. Not an invoice or transaction rate.recentObserved 2026-09-09T22:35:30Z · valid through 2026-09-10T04:35:30ZOfficial source ↗
Microsoft AzureA100 · SXM 40GBeastuson demand · 8 GPU/node$3.3996recentObserved 2026-09-09T22:35:25Z · valid through 2026-09-10T04:35:25ZOfficial source ↗
CivoH200 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$3.4900recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
ScalewayH100-SXM · 80GBfr-par-2on demand · 8 GPU/node$3.6894Indicative SaaS Sentinel calculation. Native price €25.3308/node-hour ( €3.1663/GPU-hour). USD uses 1.1652 USD/EUR from the ECB reference dated 2026-09-09. ECB reference rate ↗. ECB reference data are available free from the ECB. Not an invoice or transaction rate.recentObserved 2026-09-09T22:35:30Z · valid through 2026-09-10T04:35:30ZOfficial source ↗
LambdaH100 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$3.7900recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
Nebius AI CloudH100 · NVLink 80GBeu-north1on demand · 1 GPU/node$3.8500recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
Crusoe CloudH100 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$3.9000recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
Nebius AI CloudB200 · NVLink 180GBme-west1spot · 1 GPU/node$3.9500recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
HyperstackH200 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$3.9900recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
DataCrunch (now Verda)H200 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$4.0000recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
Crusoe CloudH200 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$4.2900recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
Nebius AI CloudB300 · NVLink 288GBeu-west2spot · 1 GPU/node$4.3000recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
DigitalOceanH100 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$4.4100recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
DigitalOceanH200 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$4.4700recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
Nebius AI CloudH200 · NVLink 141GBeu-north1on demand · 1 GPU/node$4.5000recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
PaperspaceH100 · SXM 80GBny2on demand · 1 GPU/node$5.9500recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
VerdaB200 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$6.1100recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
LambdaB200 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$6.6900recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
Microsoft AzureH100 · NVL 94GBeastuson demand · 2 GPU/node$6.9800recentObserved 2026-09-09T22:35:25Z · valid through 2026-09-10T04:35:25ZOfficial source ↗
Nebius AI CloudB200 · NVLink 180GBme-west1on demand · 1 GPU/node$7.1500recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
Nebius AI CloudB300 · NVLink 288GBeu-west2on demand · 1 GPU/node$7.8500recentObserved 2026-09-09T22:35:29Z · valid through 2026-09-10T22:35:29ZOfficial source ↗
CoreWeaveB200 · Variant not specifiedGlobal / unspecifiedon demand · 1 GPU/node$8.6000recentObserved 2026-09-09T22:35:19Z · valid through 2026-09-10T22:35:19ZOfficial source ↗
ScalewayB300-SXM · 288GBfr-par-2on demand · 8 GPU/node$8.7390Indicative SaaS Sentinel calculation. Native price €60.0000/node-hour ( €7.5000/GPU-hour). USD uses 1.1652 USD/EUR from the ECB reference dated 2026-09-09. ECB reference rate ↗. ECB reference data are available free from the ECB. Not an invoice or transaction rate.recentObserved 2026-09-09T22:35:30Z · valid through 2026-09-10T04:35:30ZOfficial source ↗

Before you commit

Test this estimate against the costs the calculator cannot know

A buyer-funded review challenges your assumptions against redacted quotes, capacity evidence, contract terms and excluded costs without changing the technical ranking.

Transparent calculation

What the result includes—and what it does not

Monthly spend

official GPU-hour rate × billable hours

The selected catalog rate is already normalized to one GPU-hour. A provider may bill a larger node or impose other minimums; check its source.

Effective output

tokens/sec × utilization × 3,600 × hours

Throughput and utilization come entirely from you. They should reflect output tokens, not prompts, and the same workload definition across scenarios.

Cost per million

monthly GPU cost ÷ monthly tokens × 1,000,000

Hours cancel mathematically in this unit cost, but remain visible because they determine monthly spend and token volume.

Excluded: CPUs bundled with a node, storage, network egress, orchestration, model hosting software, support, taxes, idle capacity outside the utilization estimate, failed requests, prompt-token cost and provider availability. This is a planning calculation, not a performance benchmark or purchasing claim.