SaaS SentinelGPU decision assuranceData status

Data sources

Published GPU pricing and benchmark sources

This index lists the pricing and workload evidence used by the public calculators so readers can verify both the hourly rate and the measured result.

Open global databaseBrowse calculatorsReport a source issue

Source-use governance

Retrieval, comparison and redistribution are separate gates

Technical access does not automatically grant permission to reproduce provider content, redistribute provider-derived fields in bulk or resell them. The registry was reviewed on 2026-08-13 and fails closed after 2026-09-12 unless it is reviewed again.

Automated retrieval

5 documented APIs are enabled. 10 official public pages remain conditional on their published terms and minimal attributed use.

Downstream rights

The site licenses its original selection, normalization, calculations and presentation only. Provider content reproduction, bulk redistribution and paid resale are not granted without provider authorization.

Fail-closed suspension

1 provider source is suspended by policy. Historical rows remain auditable but cannot enter current comparisons or automated refreshes.

Suspended source · runpod

Automated access disabled

Terms of Service section 4, automated or non-human access restriction. Reactivation requires dated written permission or qualified legal review.

Open provider terms

Official benchmark evidence

MLCommons MLPerf 6.0 repositories

Benchmark measurements are versioned separately from pricing. The public dataset pins both repositories to full commits and retains SHA-256 hashes for the summary and every raw training run.

Inference · pinned 4d3916ac9cf4

MLPerf Inference v6.0

Closed-division, available datacenter results for the exact Nebius 8× B300 submission.

Open official repository

Training · pinned eabf23a07b2a

MLPerf Training v6.0

Twenty successful raw runs across two time-to-quality workloads, aggregated with the documented trimmed mean.

Open official repository

Inspect published benchmark economicsOpen benchmark JSON

Canonical catalog sources

Structured and official inputs

These are the source groups feeding the normalized global database. Large APIs are summarized once instead of repeated for every regional record.

official api

Akamai Cloud Computing

Normalized records
429
Named regions
33
GPU models
RTX 4000 Ada, RTX 6000
Latest observation
2026-08-13
Automation
enabled documented api
Current comparison
enabled with attribution

Freshness states: recent.

Open official source · Provider terms · Inspect provider offers

official html

Paperspace

Normalized records
27
Named regions
3
GPU models
A100, A4000, A5000, A6000, H100, P4000, P5000, P6000, RTX 4000, RTX 5000, V100
Latest observation
2026-08-13
Automation
conditional public page
Current comparison
enabled with attribution

Freshness states: recent.

Open official source · Provider terms · Inspect provider offers

official html

Nebius AI Cloud

Normalized records
20
Named regions
7
GPU models
B200, B300, H100, H200, RTX PRO 6000
Latest observation
2026-08-13
Automation
conditional public page
Current comparison
enabled with attribution

Freshness states: recent.

Open official source · Provider terms · Inspect provider offers

official html

DigitalOcean

Normalized records
4
Named regions
0
GPU models
H100, H200, L40S, RTX 6000 Ada
Latest observation
2026-08-13
Automation
conditional public page
Current comparison
enabled with attribution

Freshness states: recent.

Open official source · Provider terms · Inspect provider offers

official html

DataCrunch (now Verda)

Normalized records
2
Named regions
0
GPU models
H200, RTX 6000 Ada
Latest observation
2026-08-13
Automation
conditional public page
Current comparison
enabled with attribution

Freshness states: recent.

Open official source · Provider terms · Inspect provider offers

Providers covered

15 current providers; 16 retained in the audit catalog

Current comparison: Akamai Cloud Computing, Civo, CoreWeave, Crusoe Cloud, DataCrunch (now Verda), DigitalOcean, Hyperstack, Lambda, Microsoft Azure, Nebius AI Cloud, OVHcloud, Paperspace, Scaleway, Verda, Vultr.

Audit-only providers remain visible with an explicit historical or suspended state; they are not counted as current market coverage.

Source index

Pricing references used by calculators

The calculator link shows where the source is used. The external link opens the provider or documentation page used for the published rate.