Sources and update dates
Rental prices come from providers, hardware specifications from manufacturers, model parameters from publisher records, and purchase prices from named retailers. Competitor price tables are not sources.
GPU rental prices
Checked dates belong to each provider. A recent page build does not renew an older observation. Rates are USD per GPU-hour, before taxes; a quoted node total is divided by its published GPU count.
| Provider | Official pricing source | Checked | Price check |
|---|---|---|---|
| Daytona | billing.app.daytona.io | 21 November 2026 | Prices read |
| Lium | lium.io | 21 November 2026 | Prices read |
| Impossible Cloud | www.impossiblecloud.com | 21 November 2026 | Prices read |
| Synteq | synteq.com | 21 November 2026 | Prices read |
| HyperAI | hyper.ai | 21 November 2026 | Prices read |
| Runcrate | www.runcrate.ai | 21 November 2026 | Prices read |
| GPUwerk | gpuwerk.com | 21 November 2026 | Prices read |
| RunPod | www.runpod.io | 21 November 2026 | Prices read |
| Lambda | lambda.ai | 21 November 2026 | Prices read |
| Hyperstack | www.hyperstack.cloud | 21 November 2026 | Prices read |
| CoreWeave | www.coreweave.com | 21 November 2026 | Prices read |
| Nebius | docs.nebius.com | 21 November 2026 | Prices read |
| Verda (DataCrunch) | datacrunch.io | 21 November 2026 | Prices read |
| Voltage Park | support.voltagepark.com | 21 November 2026 | Prices read |
| Crusoe | www.crusoe.ai | 21 November 2026 | Prices read |
| Jarvislabs | jarvislabs.ai | 21 November 2026 | Prices read |
| OVHcloud | www.ovhcloud.com | 21 November 2026 | Prices read |
| Spheron | www.spheron.network | 21 November 2026 | Prices read |
| Microsoft Azure | prices.azure.com | 21 November 2026 | Prices read |
| AWS | b0.p.awsstatic.com | 21 November 2026 | Prices read |
| Oracle Cloud | apexapps.oracle.com | 21 November 2026 | Prices read |
| Hetzner | www.hetzner.com | 21 November 2026 | Prices read |
| Together AI | www.together.ai | 21 November 2026 | Prices read |
| fal | fal.ai | 21 November 2026 | Prices read |
| Modal | modal.com | 21 November 2026 | Prices read |
| Vultr | www.vultr.com | 21 November 2026 | Prices read |
| Massed Compute | massedcompute.com | 21 November 2026 | Prices read |
| TensorDock | www.tensordock.com | 21 November 2026 | Prices read |
| SaladCloud | salad.com | 21 November 2026 | Prices read |
| Replicate | replicate.com | 21 November 2026 | Prices read |
| Fireworks AI | fireworks.ai | 21 November 2026 | Prices read |
| Baseten | www.baseten.co | 21 November 2026 | Prices read |
| DigitalOcean | www.digitalocean.com | 21 November 2026 | Prices read |
| GMI Cloud | www.gmicloud.ai | 21 November 2026 | Prices read |
| Prime Intellect | www.primeintellect.ai | 31 October 2026 | Earlier prices kept, with their dates; the latest price check failed |
| Cerebrium | www.cerebrium.ai | 21 November 2026 | Prices read |
| Koyeb | www.koyeb.com | 21 November 2026 | Prices read |
| Thunder Compute | www.thundercompute.com | 21 November 2026 | Prices read |
| Latitude.sh | www.latitude.sh | 21 November 2026 | Prices read |
| Hot Aisle | hotaisle.xyz | 21 November 2026 | Prices read |
| Scaleway | www.scaleway.com | 21 November 2026 | Prices read |
| Google Cloud | cloud.google.com | 21 November 2026 | Prices read |
Availability is reported only when the source explicitly publishes stock. “Not reported” means there is no stock observation, even when the price is current. Reserved rates carry a term, minimum volume or an explicit statement that the commitment was not published. Monthly rates use 720 hours; annual Latitude plans use 8,760 hours. Those conversions do not establish payment timing or cancellation rights.
Retired price sources
SF Compute public price history ends with the last public observation on 24 September 2026. Prices have not been public since 25 September 2026. Collection ended on 4 October 2026; this source is excluded from current provider counts.
Hardware specifications
Specifications checked on 18 July 2026, unless a later date is given. Manufacturer specifications describe the named variant, not measured application performance. A peak compute figure is not a benchmark. Unspecified attributes are left out.
- H100 SXM: manufacturer specification
- H100 PCIe: manufacturer specification
- H100 NVL: manufacturer specification
- H200 (SXM): manufacturer specification
- H200 NVL: manufacturer specification, checked 16 October 2026
- B200: manufacturer specification
- B300: manufacturer specification
- GB300: manufacturer specification
- GB200: manufacturer specification
- H20: manufacturer specification
- GH200: manufacturer specification
- A100 SXM 80GB: manufacturer specification
- A100 PCIe 80GB: manufacturer specification
- A100 SXM 40GB: manufacturer specification
- A100 PCIe 40GB: manufacturer specification
- MI355X: manufacturer specification
- MI350X: manufacturer specification, checked 15 October 2026
- MI325X: manufacturer specification
- MI300X: manufacturer specification
- L40S: manufacturer specification
- L40: manufacturer specification
- L4: manufacturer specification
- A40: manufacturer specification
- A30: manufacturer specification
- A10: manufacturer specification
- Tesla V100: manufacturer specification
- Tesla V100 SXM2: manufacturer specification, checked 15 October 2026
- Tesla V100 PCIe: manufacturer specification, checked 15 October 2026
- Tesla T4: manufacturer specification
- RTX 6000 Ada: manufacturer specification
- RTX A6000: manufacturer specification
- RTX A5000: manufacturer specification
- RTX A4000: manufacturer specification
- RTX A2000: manufacturer specification, checked 22 October 2026
- RTX A4500: manufacturer specification, checked 17 October 2026
- RTX 5000 Ada: manufacturer specification, checked 17 October 2026
- RTX 4000 Ada: manufacturer specification, checked 17 October 2026
- RTX 4000 SFF Ada: manufacturer specification, checked 17 October 2026
- RTX 2000 Ada: manufacturer specification, checked 17 October 2026
- Quadro RTX 6000: manufacturer specification
- Quadro RTX 5000: manufacturer specification
- RTX PRO 6000: manufacturer specification, checked 16 October 2026
- RTX PRO 6000 Workstation: manufacturer specification, checked 16 October 2026
- RTX PRO 6000 Server: manufacturer specification, checked 15 October 2026
- RTX PRO 6000 Max-Q: manufacturer specification, checked 15 October 2026
- RTX PRO 4000 Blackwell: manufacturer specification, checked 15 October 2026
- RTX PRO 4500 Blackwell: manufacturer specification, checked 15 October 2026
- RTX PRO 5000 Blackwell: manufacturer specification, checked 15 October 2026
- RTX 5090: manufacturer specification
- RTX 5080: manufacturer specification
- RTX 4090: manufacturer specification
- RTX 3090: manufacturer specification, checked 17 October 2026
- RTX 3090 Ti: manufacturer specification, checked 17 October 2026
- RTX 5070 Ti: manufacturer specification, checked 16 October 2026
- RTX 5070: manufacturer specification, checked 16 October 2026
- RTX 4070: manufacturer specification, checked 16 October 2026
- RTX 5060 Ti 16GB: manufacturer specification, checked 16 October 2026
- RTX 5060: manufacturer specification, checked 16 October 2026
- RTX 4070 Ti Super: manufacturer specification, checked 16 October 2026
- RTX 4060 Ti 16GB: manufacturer specification, checked 16 October 2026
- RTX 4080: manufacturer specification, checked 16 October 2026
- RTX 4070 Ti: manufacturer specification, checked 16 October 2026
- RTX 3080: manufacturer specification, checked 16 October 2026
- RTX 3070: manufacturer specification, checked 16 October 2026
- RTX 4080 SUPER: manufacturer specification, checked 17 October 2026
- RTX 3080 Ti: manufacturer specification, checked 17 October 2026
- RTX 5090 Laptop: manufacturer specification; memory bandwidth: NVIDIA's RTX 50 Series laptops page; the desktop RTX 5090's 3,352 AI TOPS: NVIDIA's RTX 5090 page, checked 23 October 2026
Company facts
- Lambda, Founded: official company page. Checked 27 September 2026.
- Lambda, Account access: official company page. Checked 27 September 2026.
- Lambda, GPU regions: official company page. Checked 18 October 2026.
- RunPod, Published compliance: official company page. Checked 27 September 2026.
- RunPod, Named customer example: official company page. Checked 27 September 2026.
- RunPod, Data centres: official company page. Checked 18 October 2026.
- RunPod, Data centre in Japan: official company page. Checked 18 October 2026.
- Nebius, Headquarters: official company page. Checked 27 September 2026.
- Nebius, GPU regions: official company page. Checked 18 October 2026.
- CoreWeave, Founded: official company page. Checked 17 October 2026.
- CoreWeave, Data centres: official company page. Checked 17 October 2026.
- Hot Aisle, Founded: official company page. Checked 17 October 2026.
- Hot Aisle, Data centre: official company page. Checked 18 October 2026.
- DigitalOcean, Founded: official company page. Checked 17 October 2026.
- DigitalOcean, GPU data centres: official company page. Checked 18 October 2026.
- Hyperstack, Operator: official company page. Checked 17 October 2026.
- Hyperstack, Data centre regions: official company page. Checked 17 October 2026.
- Modal, Team locations: official company page. Checked 17 October 2026.
- Modal, Container regions: official company page. Checked 18 October 2026.
- fal, Founded: official company page. Checked 17 October 2026.
- fal, Serverless regions: official company page. Checked 17 October 2026.
- SaladCloud, Founded: official company page. Checked 17 October 2026.
- SaladCloud, Where most nodes are: official company page. Checked 18 October 2026.
- SaladCloud, GPU nodes: official company page. Checked 18 October 2026.
- Baseten, Founded: official company page. Checked 17 October 2026.
- Baseten, GPU regions: official company page. Checked 18 October 2026.
- Cerebrium, Founded: official company page. Checked 17 October 2026.
- Cerebrium, GPU regions: official company page. Checked 18 October 2026.
- Verda (DataCrunch), Founded: official company page. Checked 17 October 2026.
- Verda (DataCrunch), Current name: official company page. Checked 17 October 2026.
- Verda (DataCrunch), Data centres: official company page. Checked 17 October 2026.
- Scaleway, Operator: official company page. Checked 17 October 2026.
- Vast.ai, Founded: official company page. Checked 17 October 2026.
- OVHcloud, Founded: official company page. Checked 17 October 2026.
- OVHcloud, GPU locations: official company page. Checked 18 October 2026.
- Crusoe, Founded: official company page. Checked 17 October 2026.
- Crusoe, Regions: official company page. Checked 18 October 2026.
- Vultr, Founded: official company page. Checked 17 October 2026.
- Koyeb, Founded: official company page. Checked 17 October 2026.
- Koyeb, GPU locations: official company page. Checked 18 October 2026.
- Thunder Compute, Founded: official company page. Checked 17 October 2026.
- Thunder Compute, Machines in: official company page. Checked 18 October 2026.
- Massed Compute, Founded: official company page. Checked 17 October 2026.
- Massed Compute, Data centres: official company page. Checked 17 October 2026.
- Lium, Operator: official company page. Checked 17 October 2026.
- Impossible Cloud, Founded: official company page. Checked 17 October 2026.
- Impossible Cloud, Headquarters: official company page. Checked 17 October 2026.
- Replicate, Owner: official company page. Checked 17 October 2026.
- Replicate, Existing APIs: official company page. Checked 17 October 2026.
- Synteq, Infrastructure: official company page. Checked 17 October 2026.
- Synteq, GPU data centres: official company page. Checked 18 October 2026.
- Voltage Park, Owner: official company page. Checked 17 October 2026.
- Voltage Park, Existing contracts: official company page. Checked 17 October 2026.
- Voltage Park, Infrastructure: official company page. Checked 17 October 2026.
- TensorDock, Owner: official company page. Checked 17 October 2026.
- TensorDock, Hosts: official company page. Checked 17 October 2026.
- Jarvislabs, GPU regions: official company page. Checked 17 October 2026.
- GMI Cloud, Infrastructure: official company page. Checked 17 October 2026.
- Together AI, Founders: official company page. Checked 17 October 2026.
- Together AI, GPU cluster locations: official company page. Checked 18 October 2026.
- Latitude.sh, Infrastructure: official company page. Checked 17 October 2026.
- Prime Intellect, How it sources GPUs: official company page. Checked 17 October 2026.
- Spheron, How it sources GPUs: official company page. Checked 17 October 2026.
- Spheron, Capacity: official company page. Checked 17 October 2026.
- HyperAI, Company: official company page. Checked 17 October 2026.
- AWS, Regions: official company page. Checked 17 October 2026.
- Microsoft Azure, Regions: official company page. Checked 17 October 2026.
- Google Cloud, Regions: official company page. Checked 17 October 2026.
- Google Cloud, GPU zones: official company page. Checked 18 October 2026.
- Fireworks AI, Founded: official company page. Checked 17 October 2026.
- Fireworks AI, Latest funding: official company page. Checked 17 October 2026.
- Fireworks AI, GPU regions: official company page. Checked 17 October 2026.
- Daytona, Founded: official company page. Checked 17 October 2026.
- Daytona, Latest funding: official company page. Checked 17 October 2026.
- Daytona, Regions: official company page. Checked 17 October 2026.
- Runcrate, Operator: official company page. Checked 31 October 2026.
- Runcrate, How it sources GPUs: official company page. Checked 31 October 2026.
- GPUwerk, Operator: official company page. Checked 31 October 2026.
- GPUwerk, Data centre: official company page. Checked 31 October 2026.
- GPUwerk, How it sources GPUs: official company page. Checked 31 October 2026.
- Oracle Cloud, Operator: official company page. Checked 3 November 2026.
- Oracle Cloud, Regions: official company page. Checked 3 November 2026.
- Oracle Cloud, Pricing by region: official company page. Checked 3 November 2026.
- Hetzner, Operator: official company page. Checked 3 November 2026.
- Hetzner, GPU server locations: official company page. Checked 3 November 2026.
Rubin sources and specifications
Specifications checked on 5 October 2026. GPU figures refer to NVIDIA's individual-GPU table, not a cloud instance or a complete rack. Peak compute specifications are not measured workload performance.
| Claim | Official source | Quoted evidence |
|---|---|---|
| Rubin individual GPU specifications | www.nvidia.com | 288 GB HBM4; 22 TB/s; 3.6 TB/s; 4 PFLOPS; 50 PFLOPS |
| Vera Rubin NVL72 | www.nvidia.com | 72 NVIDIA Rubin GPUs |
| B200 GPU memory | docs.nvidia.com | B200 180 GB product |
| B300 GPU memory | docs.nvidia.com | 8 x 288 GB = 2.3 TB total |
| H200 SXM GPU memory | www.nvidia.com | 141GB |
| B200 memory type | www.nvidia.com | 1,440 GB total, 64 TB/s HBM3e bandwidth |
| B300 memory type | www.nvidia.com | HBM3E |
| NVIDIA launch and cloud partners, 5 January 2026 | nvidianews.nvidia.com | CoreWeave, Lambda, Nebius and Nscale. |
| CoreWeave production announcement | www.coreweave.com | announced the availability of NVIDIA Vera Rubin NVL72 on CoreWeave |
| Lambda Rubin product page | lambda.ai | Available H2 2026. |
| Spheron pre-order listing | www.spheron.network | Ships H2 2026 |
NVIDIA publishes two sets of per-GPU Rubin figures. Its HGX table lists 22 TB/s of memory bandwidth and 3.6 TB/s of NVLink bandwidth; its Vera Rubin NVL72 table lists 19.2 TB/s and 3 TB/s, for an at-scale AI factory. Ask the provider which system it rents before comparing performance.
Provider fees, access and compliance
Checked 7 July 2026; cells rechecked later show their own date on the fees and enterprise matrices. These are documentation observations, not an audit of a provider. Conditional rates and unpublished fields cannot be turned into a complete bill.
- Hetzner: www.hetzner.com/dedicated-rootserver/matrix-gpu/
- Oracle Cloud: www.oracle.com/cloud/price-list/ · www.oracle.com/cloud/compute/faq/ · docs.oracle.com/en-us/iaas/releasenotes/changes/20facdce-34bf-4648-a0a7-de5db73657ce/
- RunPod: docs.runpod.io/pods/overview · docs.runpod.io/pods/pricing · docs.runpod.io/accounts-billing/billing · www.runpod.io/enterprise · www.runpod.io/legal/compliance · www.runpod.io/legal/data-processing-agreement · docs.runpod.io/references/billing-information · trust.runpod.io/ · docs.runpod.io/pods/templates/overview · docs.runpod.io/serverless/pricing
- Lambda: docs.lambda.ai/public-cloud/billing/ · docs.lambda.ai/public-cloud/manage-billing/ · lambda.ai/pricing · trust.lambda.ai/ · docs.lambda.ai/public-cloud/1-click-clusters/support/ · docs.lambda.ai/public-cloud/on-demand/ · lambda.ai/instances
- Hyperstack: www.hyperstack.cloud/gpu-pricing · docs.hyperstack.cloud/docs/faq/ · docs.hyperstack.cloud/docs/billing/billing-policies/ · www.hyperstack.cloud/slas · www.hyperstack.cloud/secure-by-design · www.hyperstack.cloud/data-processing · docs.hyperstack.cloud/docs/billing/billing-policies · www.hyperstack.cloud/
- CoreWeave: www.coreweave.com/pricing · docs.coreweave.com/platform/capacity-plans.md · docs.coreweave.com/policies/terms-of-service · trust.coreweave.com/ · docs.coreweave.com/policies/terms-of-service/data-processing-agreement · www.coreweave.com/direct-to-expert-support · www.coreweave.com/ai-data-centers · www.coreweave.com/platform
- Nebius: nebius.com/prices · docs.nebius.com/compute/resources/pricing · docs.nebius.com/legal/sla-levels/compute · nebius.com/trust-center/soc-2 · nebius.com/trust-center/iso-27001 · docs.nebius.com/legal/hipaa · docs.nebius.com/legal/dpa · docs.nebius.com/signup-billing/payments/bank-transfers · docs.nebius.com/overview/support · docs.nebius.com/overview/regions · docs.nebius.com/
- Verda (DataCrunch): verda.com/pricing · docs.verda.com/cpu-and-gpu-instances/shutdown-hibernate-and-delete/ · docs.verda.com/welcome-to-verda/pricing-and-billing/ · verda.com/ai-cloud · verda.com/terms-and-conditions · docs.verda.com/welcome-to-verda/contact-support/ · docs.verda.com/welcome-to-verda/locations-and-sustainability/
- Voltage Park: www.voltagepark.com/pricing · support.voltagepark.com/article/fa-qs · www.voltagepark.com/computer-services-agreement · www.voltagepark.com/virtual-machine-sla · www.voltagepark.com/security · trust.voltagepark.com/ · www.voltagepark.com/enterprise · www.voltagepark.com/about · www.voltagepark.com/ · www.voltagepark.com/infrastructure
- Crusoe: www.crusoe.ai/cloud/pricing · docs.crusoecloud.com/storage/disks/overview · docs.crusoecloud.com/compute/virtual-machines/overview · legal.crusoe.ai/ · www.crusoe.ai/cloud · support.crusoecloud.com/hc/en-us · docs.crusoecloud.com/locations.md · www.crusoe.ai/
- Jarvislabs: jarvislabs.ai/pricing · docs.jarvislabs.ai/object-storage/ · docs.jarvislabs.ai/faqs/ · jarvislabs.ai/ · jarvislabs.ai/termsandservice · docs.jarvislabs.ai/getting_started/
- OVHcloud: us.ovhcloud.com/public-cloud/prices/ · us.ovhcloud.com/public-cloud/network/ · docs.ovhcloud.com/en/guides/public-cloud/compute/suspend-or-pause-an-instance · us.ovhcloud.com/public-cloud/faq/ · www.ovhcloud.com/en/public-cloud/gpu/ · www.ovhcloud.com/en/compliance/soc-1-2-3/ · www.ovhcloud.com/en/compliance/iso-27001-27017-27018/ · us.ovhcloud.com/legal/data-processing-agreement/ · support.us.ovhcloud.com/hc/en-us/articles/22179651508115-Public-Cloud-billing · us.ovhcloud.com/support-levels/ · www.ovhcloud.com/en/public-cloud/regions-availability/ · us.ovhcloud.com/public-cloud/
- Spheron: www.spheron.network/pricing/ · docs.spheron.ai/billing · www.spheron.network/terms · www.spheron.network/ · docs.spheron.network/overview
- Microsoft Azure: azure.microsoft.com/en-us/pricing/details/bandwidth/ · azure.microsoft.com/api/v2/pricing/managed-disks/calculator/ · learn.microsoft.com/en-us/azure/virtual-machines/states-billing · learn.microsoft.com/en-us/azure/virtual-machines/vm-usage · www.microsoft.com/licensing/docs/documents/download/OnlineSvcsConsolidatedSLA(WW)(English)(October_2026)(CR).docx · learn.microsoft.com/en-us/compliance/regulatory/offering-soc-2 · learn.microsoft.com/en-us/compliance/regulatory/offering-iso-27001 · learn.microsoft.com/en-us/compliance/regulatory/offering-hipaa-hitech · www.microsoft.com/licensing/docs/view/Microsoft-Products-and-Services-Data-Protection-Addendum-DPA · learn.microsoft.com/en-us/azure/cost-management-billing/manage/pay-by-invoice · azure.microsoft.com/en-us/support/plans/ · azure.microsoft.com/en-us/explore/global-infrastructure/geographies/ · azure.microsoft.com/en-us/products/machine-learning/ · azure.microsoft.com/en-us/products/virtual-machines/ · azure.microsoft.com/en-us/pricing/spot/ · azure.microsoft.com/en-us/pricing/reserved-vm-instances/
- AWS: b0.p.awsstatic.com/pricing/2.0/meteredUnitMaps/datatransfer/USD/current/datatransfer.json · aws.amazon.com/ebs/pricing/ · docs.aws.amazon.com/AWSEC2/latest/UserGuide/how-ec2-instance-stop-start-works.html · aws.amazon.com/ec2/pricing/ · aws.amazon.com/compute/sla/ · aws.amazon.com/blogs/security/fall-2022-soc-2-type-2-privacy-report-now-available/ · aws.amazon.com/compliance/iso-27001-faqs/ · aws.amazon.com/compliance/hipaa-compliance/ · aws.amazon.com/compliance/gdpr-center/ · docs.aws.amazon.com/awsaccountbilling/latest/aboutv2/manage-payment-method.html · aws.amazon.com/premiumsupport/plans/ · aws.amazon.com/about-aws/global-infrastructure/regions_az/ · aws.amazon.com/sagemaker/notebooks/ · aws.amazon.com/ec2/ · aws.amazon.com/ec2/spot/ · aws.amazon.com/ec2/pricing/reserved-instances/
- Together AI: docs.together.ai/docs/gpu-clusters-billing.md · www.together.ai/pricing · docs.together.ai/docs/billing-credits.md · www.together.ai/gpu-clusters · docs.together.ai/docs/billing · www.together.ai/support
- fal: fal.ai/docs/documentation/serverless/pricing · fal.ai/learn/serverless/ai-inference-provider-for-generative-media · fal.ai/legal/data-processing-addendum · fal.ai/terms · fal.ai/ · fal.ai/docs/documentation/compute · fal.ai/pricing
- Modal: modal.com/docs/guide/network-egress-billing · modal.com/pricing · modal.com/docs/guide/billing · modal.com/docs/guide/security · modal.com/blog/hipaa · modal.com/legal/dpa · modal.com/docs/guide/region-selection · modal.com/docs/guide/notebooks
- Vultr: docs.vultr.com/support/platform/billing/what-is-the-bandwidth-overage-rate · docs.vultr.com/support/platform/billing/how-is-bandwidth-usage-calculated · docs.vultr.com/support/platform/billing/how-is-object-storage-billed · docs.vultr.com/support/platform/billing/are-stopped-instances-still-billed-on-vultr · docs.vultr.com/support/platform/billing/how-am-i-billed-for-my-servers · www.vultr.com/legal/sla/ · www.vultr.com/trust-center/ · www.vultr.com/legal/compliance/ · docs.vultr.com/support/platform/billing/how-do-i-arrange-payment-via-ach-or-wire · www.vultr.com/support/ · www.vultr.com/products/cloud-gpu/ · www.vultr.com/api/
- Massed Compute: vm.massedcompute.com/pricing · api-docs.massedcompute.com/api/v1 · massedcompute.com/products/on-demand/ · massedcompute.com/on-demand/ · massedcompute.com/ · massedcompute.com/dpa/ · vm-docs.massedcompute.com/docs/billing/overview · vm-docs.massedcompute.com/ · vm-docs.massedcompute.com/docs/intro
- Google Cloud: cloud.google.com/vpc/network-pricing · cloud.google.com/compute/disks-image-pricing · cloud.google.com/compute/docs/instances/instance-lifecycle · cloud.google.com/products/compute/pricing
- DigitalOcean: docs.digitalocean.com/products/droplets/details/pricing/ · docs.digitalocean.com/products/volumes/details/pricing/ · www.digitalocean.com/pricing/gpu-droplets
- TensorDock: tensordock.com/ · docs.tensordock.com/virtual-machines/spot-instances · www.tensordock.com/gpu-h100.html · console.tensordock.com/deploy · dashboard.tensordock.com/deploy
- Thunder Compute: www.thundercompute.com/pricing · www.thundercompute.com/docs/guides/stopping-instances · www.thundercompute.com/docs/billing
- Latitude.sh: www.latitude.sh/docs/usage/bandwidth · www.latitude.sh/pricing/storage · www.latitude.sh/docs/vms/virtual-machines · www.latitude.sh/docs/billing/payment-methods · www.latitude.sh/docs/billing/on-demand-vs-reserved
- Scaleway: www.scaleway.com/en/gpu-instances/ · www.scaleway.com/en/pricing/storage/ · www.scaleway.com/en/docs/instances/faq/
- Hot Aisle: hotaisle.xyz/quick-start/ · hotaisle.xyz/pricing/ · hotaisle.xyz/
- Prime Intellect: docs.primeintellect.ai/api-reference/managing-disks · docs.primeintellect.ai/faq
- SaladCloud: salad.com/pricing/ · docs.salad.com/general/explanation/billing
- Replicate: replicate.com/docs/topics/billing · replicate.com/docs/topics/billing/prepaid-credit · replicate.com/pricing
- Fireworks AI: docs.fireworks.ai/guides/ondemand-deployments · docs.fireworks.ai/guides/quotas_usage/account-quotas · fireworks.ai/pricing
- Baseten: docs.baseten.co/organization/billing
- GMI Cloud: docs.gmicloud.ai/cluster-engine/user-management/account · www.gmicloud.ai/en/pricing
- Cerebrium: cerebrium.ai/pricing · cerebrium.ai/docs/calculating-cost
- Koyeb: www.koyeb.com/pricing · www.koyeb.com/docs/faqs/pricing
- Daytona: www.daytona.io/pricing · www.daytona.io/docs/en/billing/ · www.daytona.io/docs/en/limits/
- Lium: docs.lium.io/pod-users/volumes · docs.lium.io/pod-users/billing · docs.lium.io/pod-users/faq
- Impossible Cloud: www.impossiblecloud.com/pricing · www.impossiblecloud.com/baremetal
- Synteq: synteq.com/compare/products/gpu · synteq.com/compare/products/storage · cloud.synteq.com/docs/refund-eligibility
- HyperAI: hyper.ai/en/pricing · hyper.ai/en/docs/getting-started/quota-and-usage
- Runcrate: www.runcrate.ai/docs/faq/billing · www.runcrate.ai/pricing · www.runcrate.ai/docs/billing/pricing
- GPUwerk: gpuwerk.com/pricing/ · gpuwerk.com/availability.txt
Model parameter counts and memory
Parameter counts come from the publisher’s Safetensors metadata. For standard attention, cache bytes = 2 × layers × key/value heads × head dimension × context tokens × batch size × 2 bytes. Weight bytes = parameters × precision bits ÷ 8. Memory tables use decimal GB, with editable overhead assumptions; they do not guarantee that a particular runtime will load a model.
The full-training scenario uses the Transformers mixed-precision AdamW memory accounting: 18 bytes per parameter, plus activations supplied by the reader. Adapter training applies that accounting only to the assumed trainable parameter fraction. These are budget estimates, not throughput measurements.
LLM API rates
Serverless billing
Runpod billing documentation. Checked 18 October 2026. Startup, execution and idle timeout are billed, with whole-second rounding; active-worker discounts require a sales inquiry.
Purchase prices
- A10 24GB: NetworkOutlet. Checked 21 November 2026.
- A100 40GB PCIe: NetworkOutlet. Checked 21 November 2026.
- A100 80GB SXM4 module: NetworkOutlet. Checked 21 November 2026.
- A100 80GB PCIe: SabrePC. Checked 21 November 2026.
- A100 80GB PCIe: NetworkOutlet. Checked 21 November 2026.
- A30 24GB: Viperatech. Checked 21 November 2026.
- A40 48GB: ServerPartDeals. Checked 21 November 2026.
- A40 48GB: Viperatech. Checked 21 November 2026.
- Exxact TensorEX TS4-142487900 8U server, 8× B200 180GB (AMD EPYC), Essential configuration: Exxact. Checked 21 November 2026.
- Exxact TensorEX TS4-154079096 10U server, 8× B200 180GB (AMD EPYC), Essential configuration: Exxact. Checked 21 November 2026.
- Exxact TensorEX TS4-169219634 10U server, 8× B200 180GB (Intel Xeon 6900), Essential configuration: Exxact. Checked 21 November 2026.
- Exxact TensorEX TS4-128275176 8U server, 8× B200 180GB (Intel Xeon 4th/5th Gen), Essential configuration: Exxact. Checked 21 November 2026.
- Supermicro SYS-822GS-NBRT server, 8× B200 180GB: NetworkOutlet. Checked 21 November 2026.
- Supermicro SYS-A21GE-NBRT server, 8× B200: Viperatech. Checked 11 November 2026.
- GIGABYTE G893-ZD1-AAX5 server, 8× B200: Viperatech. Checked 11 November 2026.
- Supermicro SYS-822GS-NB3RT 8U server, 8× B300 288GB (Intel Xeon 6), starting configuration: AVADirect. Checked 21 November 2026.
- Exxact TensorEX TS4-126572371 8U server, 8× B300 288GB (AMD EPYC), Essential configuration: Exxact. Checked 21 November 2026.
- Exxact TensorEX TS4-131188947 8U server, 8× B300 288GB (Intel Xeon 6500P/6700E/6700P), Essential configuration: Exxact. Checked 21 November 2026.
- Exxact TensorEX TS4-198437339 8U server, 8× B300 288GB (Intel Xeon 6700E/6700P), Essential configuration: Exxact. Checked 21 November 2026.
- Supermicro SYS-822GS-NB3RT server, 8× B300 (Intel Xeon): NetworkOutlet. Checked 21 November 2026.
- Exxact TensorEX TS4-104366747 8U server, 8× B300 288GB (Intel Xeon 6500P/6700P), Essential configuration: Exxact. Checked 21 November 2026.
- Supermicro AS-8126GS-NB3RT server, 8× B300 (AMD EPYC): NetworkOutlet. Checked 21 November 2026.
- Exxact TensorEX TS4-127210561 9U server, 8× B300 288GB (Intel Xeon 6700P), Essential configuration: Exxact. Checked 21 November 2026.
- Supermicro Super AI Station ARS-511GD workstation, 1× GB300 (Grace CPU and Blackwell Ultra GPU), starting configuration: AVADirect. Checked 21 November 2026.
- Supermicro Super AI Station ARS-511GD workstation, 1× GB300 (Grace CPU and Blackwell Ultra GPU): Viperatech. Checked 11 November 2026.
- MSI XpertStation WS300 workstation, 1× GB300 252GB (Grace CPU and Blackwell Ultra GPU): SabrePC. Checked 21 November 2026.
- MSI XpertStation WS300 workstation, 1× GB300 252GB (Grace CPU and Blackwell Ultra GPU), starting configuration: AVADirect. Checked 21 November 2026.
- ASUS ExpertCenter Pro ET900N G3 workstation, 1× GB300 252GB (Grace CPU and Blackwell Ultra GPU): SabrePC. Checked 21 November 2026.
- ASUS ExpertCenter Pro ET900N G3 workstation, 1× GB300 252GB (Grace CPU and Blackwell Ultra GPU), starting configuration: AVADirect. Checked 21 November 2026.
- GIGABYTE W775-V10 workstation, 1× GB300 252GB (Grace CPU and Blackwell Ultra GPU), starting configuration: AVADirect. Checked 21 November 2026.
- GIGABYTE XH23-VG0-AAJ1 2U server, 1× GH200 96GB, starting configuration: AVADirect. Checked 21 November 2026.
- GIGABYTE H263-V11-LAW1 2U server, 4× GH200 96GB, starting configuration: AVADirect. Checked 21 November 2026.
- H100 NVL 94GB: NetworkOutlet. Checked 21 November 2026.
- H100 NVL 94GB: SabrePC. Checked 21 November 2026.
- H100 80GB PCIe: SabrePC. Checked 20 November 2026.
- HGX H100 8-GPU board (SXM5): NetworkOutlet. Checked 21 November 2026.
- Dell PowerEdge XE9680 server, 8× H100 SXM: NetworkOutlet. Checked 21 November 2026.
- Exxact TensorEX TS4-140031951 5U server, 8× H100 SXM 80GB (Intel Xeon 4th/5th Gen), Essential configuration: Exxact. Checked 21 November 2026.
- Exxact TensorEX TS4-155878578 6U server, 8× H100 SXM 80GB (Intel Xeon 4th/5th Gen), Essential configuration: Exxact. Checked 21 November 2026.
- Exxact TensorEX TS4-193475697 8U server, 8× H100 SXM 80GB (AMD EPYC), Essential configuration: Exxact. Checked 21 November 2026.
- H200 NVL 141GB: SabrePC. Checked 21 November 2026.
- H200 NVL 141GB: ASA Computers. Checked 21 November 2026.
- H200 NVL 141GB: NetworkOutlet. Checked 21 November 2026.
- GIGABYTE G593-SD1-AAX3 5U server, 8× H200 SXM 141GB (Intel Xeon), starting configuration: AVADirect. Checked 21 November 2026.
- Exxact TensorEX TS4-115801521 5U server, 8× H200 SXM 141GB (Intel Xeon 4th/5th Gen), Essential configuration: Exxact. Checked 21 November 2026.
- Exxact TensorEX TS4-127744315 6U server, 8× H200 SXM 141GB (Intel Xeon 4th/5th Gen), Essential configuration: Exxact. Checked 21 November 2026.
- Exxact TensorEX TS4-181733724 5U server, 8× H200 SXM 141GB (Intel Xeon 4th/5th Gen), Essential configuration: Exxact. Checked 21 November 2026.
- Exxact TensorEX TS4-118380266 8U server, 8× H200 SXM 141GB (AMD EPYC), Essential configuration: Exxact. Checked 21 November 2026.
- Exxact TensorEX TS4-110455529 6U server, 8× H200 SXM 141GB (AMD EPYC), Essential configuration: Exxact. Checked 21 November 2026.
- L4 24GB: SabrePC. Checked 21 November 2026.
- L40 48GB: SabrePC. Checked 21 November 2026.
- L40S 48GB: ASA Computers. Checked 21 November 2026.
- L40S 48GB: NetworkOutlet. Checked 21 November 2026.
- L40S 48GB: SabrePC. Checked 21 November 2026.
- GIGABYTE G593-SX1-AAX1 5U server, 8× MI300X 192GB (Intel Xeon), starting configuration: AVADirect. Checked 21 November 2026.
- Exxact TensorEX TS4-185328443 8U server, 8× MI300X 192GB (AMD EPYC), Essential configuration: Exxact. Checked 21 November 2026.
- Supermicro AS-8126GS-TNMR 8U server, 8× MI325X 256GB (AMD EPYC), starting configuration: AVADirect. Checked 21 November 2026.
- Exxact TensorEX TS4-149591758 6U server, 8× MI350X 288GB (AMD EPYC), Essential configuration: Exxact. Checked 21 November 2026.
- GIGABYTE G893-ZX1-AAX4 8U server, 8× MI355X 288GB (AMD EPYC), starting configuration: AVADirect. Checked 21 November 2026.
- Quadro RTX 5000 16GB: Viperatech. Checked 11 November 2026.
- Quadro RTX 6000 24GB: NetworkOutlet. Checked 21 November 2026.
- RTX 2000 Ada 16GB: SabrePC. Checked 21 November 2026.
- RTX 3080 10GB: ServerPartDeals. Checked 18 November 2026.
- RTX 3090 24GB: ServerPartDeals. Checked 21 November 2026.
- RTX 4000 Ada 20GB: SabrePC. Checked 21 November 2026.
- RTX 4060 Ti 16GB: SabrePC. Checked 21 November 2026.
- RTX 4070 12GB: SabrePC. Checked 21 November 2026.
- RTX 4070 Ti 12GB: Viperatech. Checked 21 November 2026.
- RTX 4070 Ti 12GB: SabrePC. Checked 21 November 2026.
- RTX 4070 Ti SUPER 16GB: SabrePC. Checked 21 November 2026.
- RTX 4080 16GB: SabrePC. Checked 21 November 2026.
- RTX 4080 SUPER 16GB: SabrePC. Checked 21 November 2026.
- RTX 4090 24GB: SabrePC. Checked 21 November 2026.
- RTX 5000 Ada 32GB: ASA Computers. Checked 21 November 2026.
- RTX 5000 Ada 32GB: SabrePC. Checked 21 November 2026.
- RTX 5060 8GB: SabrePC. Checked 21 November 2026.
- RTX 5060 Ti 16GB: SabrePC. Checked 21 November 2026.
- RTX 5070 12GB: SabrePC. Checked 21 November 2026.
- RTX 5070 Ti 16GB: SabrePC. Checked 21 November 2026.
- RTX 5080 16GB: SabrePC. Checked 21 November 2026.
- RTX 5090 32GB: ASA Computers. Checked 21 November 2026.
- RTX 5090 32GB: SabrePC. Checked 21 November 2026.
- RTX 5090 32GB: Viperatech. Checked 21 November 2026.
- RTX 6000 Ada 48GB: SabrePC. Checked 21 November 2026.
- RTX 6000 Ada 48GB: ASA Computers. Checked 21 November 2026.
- RTX A2000 6GB: SabrePC. Checked 21 November 2026.
- RTX A2000 12GB: SabrePC. Checked 21 November 2026.
- RTX A4000 16GB: SabrePC. Checked 21 November 2026.
- RTX A4500 20GB: SabrePC. Checked 21 November 2026.
- RTX A5000 24GB: NetworkOutlet. Checked 21 November 2026.
- RTX A6000 48GB: SabrePC. Checked 21 November 2026.
- RTX PRO 4000 Blackwell 24GB: SabrePC. Checked 21 November 2026.
- RTX PRO 4500 Blackwell 32GB: SabrePC. Checked 21 November 2026.
- RTX PRO 5000 Blackwell 48GB: ASA Computers. Checked 21 November 2026.
- RTX PRO 5000 Blackwell 48GB: SabrePC. Checked 21 November 2026.
- RTX PRO 6000 Blackwell 96GB: NetworkOutlet. Checked 21 November 2026.
- RTX PRO 6000 Blackwell 96GB: SabrePC. Checked 21 November 2026.
- T4 16GB: NetworkOutlet. Checked 21 November 2026.
- Tesla V100 32GB PCIe: ServerPartDeals. Checked 21 November 2026.
A quote-only listing does not establish a purchase price. Expired or unconfirmed prices are excluded, not assigned a new date.
B200, quote only: Lenovo ThinkSystem NVIDIA HGX B200 180GB 1000W GPU from Eton Technology. Checked 27 September 2026. NVIDIA B200 Tensor Core GPU from Eton Technology. Checked 27 September 2026. Supermicro HGX B200 8U Air-Cooled System from Eton Technology. Checked 27 September 2026. Supermicro HGX B200 4U Liquid-Cooled System from Eton Technology. Checked 27 September 2026. GIGABYTE AI-DLC-Rack NVIDIA-HGX-B200 from Eton Technology. Checked 27 September 2026. GIGABYTE AI-DLC-POD NVIDIA-HGX-B200 from Eton Technology. Checked 27 September 2026. GIGABYTE AI-AC-POD NVIDIA-HGX-B200 from Eton Technology. Checked 27 September 2026. Supermicro SRS-48UAC-B200SX from Eton Technology. Checked 27 September 2026.
GB200, quote only: Supermicro SRS-GB200-NVL72 from Eton Technology. Checked 27 September 2026.
Credits
- Google Cloud: offer terms. Checked 31 October 2026. Eligibility and expiry still apply.
- AWS: offer terms. Checked 31 October 2026. Eligibility and expiry still apply.
- Microsoft Azure: offer terms. Checked 31 October 2026. Eligibility and expiry still apply.
- Oracle Cloud: offer terms. Checked 31 October 2026. Eligibility and expiry still apply.
- OVHcloud US: offer terms. Checked 31 October 2026. Eligibility and expiry still apply.
- Modal: offer terms. Checked 31 October 2026. Eligibility and expiry still apply.
- RunPod Startup Program: offer terms. Checked 31 October 2026. Eligibility and expiry still apply.
- Google for Startups Cloud Program: offer terms. Checked 31 October 2026. Eligibility and expiry still apply.
- AWS Activate: offer terms, credit guide, promotional credit terms. Checked 31 October 2026. Eligibility and expiry still apply.
- Microsoft for Startups: offer terms, credit guide. Checked 31 October 2026. Eligibility and expiry still apply.
- NVIDIA Inception: offer terms. Checked 31 October 2026. Eligibility and expiry still apply.
- AMD Developer Cloud: offer terms. Checked 31 October 2026. Eligibility and expiry still apply.
- TensorWave (AMD MI300X): offer terms. Checked 31 October 2026. Eligibility and expiry still apply.
- Azure for Students: offer terms. Checked 31 October 2026. Eligibility and expiry still apply.
- Modal for Academics: offer terms. Checked 31 October 2026. Eligibility and expiry still apply.
Price history and corrections
History starts with the first dated observation. Daily records are append-only; identified errors are excluded or reclassified through a separate correction record. Statistical pages state whether they compare the same provider and configuration or a changing market sample.
Read the calculation methods. To report a price, use its row’s email link; it includes the provider, GPU and update date. Nothing submitted through that link is stored by this site.
Data attribution and reuse
Today's rental prices and the full daily history are offered under CC BY 4.0, with attribution to GPU Rental Prices. A fixed copy of the record's first two weeks (5 to 18 July 2026) has the identifier 10.5281/zenodo.21435394.