specification: API Commons Plans
specificationVersion: '0.1'
schema: https://raw.githubusercontent.com/api-evangelist/interface-research/main/schema/api-commons.yml#/$defs/Plans
provider: TensorDock
providerId: tensordock
created: '2026-05-25'
modified: '2026-05-25'
reconciled: true
tags:
  - GPU
  - Cloud
  - Compute
  - Marketplace
  - Hourly Pricing
description: 'Reconciled pricing for TensorDock GPU and CPU virtual machines. TensorDock operates a
  marketplace model where each independent host sets per-resource prices, so the figures below reflect
  starting (entry) hourly rates. Sources: tensordock.com and marketplace.tensordock.com/deploy.'
sources:
  - https://www.tensordock.com
  - https://marketplace.tensordock.com/deploy
  - https://docs.tensordock.com/virtual-machines/spot-instances
  - https://docs.tensordock.com/virtual-machines/introduction-to-core-compute-vms
plans:
  - id: tensordock-on-demand-gpu
    name: GPU On-Demand
    type: usage-based
    description: Per-microsecond billing for dedicated GPU virtual machines. Prices vary by GPU model
      and host; starting rates listed below.
    entries:
      - label: H100 SXM5 80GB hourly
        name: gpu_h100_sxm5
        type: metered
        metric: gpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: '2.25'
        userMultiplied: false
      - label: A100 SXM4 80GB hourly
        name: gpu_a100_sxm4_80gb
        type: metered
        metric: gpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: '0.87'
        userMultiplied: false
      - label: A100 PCIe 80GB hourly
        name: gpu_a100_pcie_80gb
        type: metered
        metric: gpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: '1.27'
        userMultiplied: false
      - label: RTX 6000 Ada 48GB hourly
        name: gpu_rtx_6000_ada
        type: metered
        metric: gpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: '0.72'
        userMultiplied: false
      - label: L40S 48GB hourly
        name: gpu_l40s
        type: metered
        metric: gpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: '0.49'
        userMultiplied: false
      - label: L4 24GB hourly
        name: gpu_l4
        type: metered
        metric: gpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: '0.24'
        userMultiplied: false
      - label: RTX A6000 48GB hourly
        name: gpu_rtx_a6000
        type: metered
        metric: gpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: '0.42'
        userMultiplied: false
      - label: RTX 5090 32GB hourly
        name: gpu_rtx_5090
        type: metered
        metric: gpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: '0.56'
        userMultiplied: false
      - label: RTX 4090 24GB hourly
        name: gpu_rtx_4090
        type: metered
        metric: gpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: '0.32'
        userMultiplied: false
      - label: RTX 3090 24GB hourly
        name: gpu_rtx_3090
        type: metered
        metric: gpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: '0.17'
        userMultiplied: false
      - label: V100 SXM3 32GB hourly
        name: gpu_v100_sxm3
        type: metered
        metric: gpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: '0.30'
        userMultiplied: false
      - label: V100 SXM2 16GB hourly
        name: gpu_v100_sxm2
        type: metered
        metric: gpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: '0.21'
        userMultiplied: false

  - id: tensordock-cpu-compute
    name: CPU Compute
    type: usage-based
    description: CPU-only virtual machines on Intel Xeon or AMD EPYC for transcoding and batch processing.
    entries:
      - label: vCPU hourly
        name: vcpu_hour
        type: metered
        metric: vcpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: '0.012'
        userMultiplied: false
      - label: RAM per GB hourly
        name: ram_gb_hour
        type: metered
        metric: gb_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: 'varies'
        userMultiplied: false
        notes: Host-set; check /api/v0/client/deploy/hostnodes for live rates.
      - label: Storage per GB hourly
        name: storage_gb_hour
        type: metered
        metric: gb_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: 'varies'
        userMultiplied: false

  - id: tensordock-spot
    name: Spot Instances
    type: bid-based
    description: Interruptible spot instances where the customer sets a bid price; lowest-priced
      eligible hostnode wins. Soft-validation endpoints let clients check whether a bid will clear
      before deploying.
    entries:
      - label: Spot bid
        name: spot_bid
        type: metered
        metric: gpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: 'variable'
        userMultiplied: false

  - id: tensordock-reserved
    name: Reserved
    type: contract
    description: Volume reservations with committed-use discounts available upon request from sales.
    entries:
      - label: Reserved capacity
        name: reserved
        type: contract
        metric: gpu
        limit: -1
        timeFrame: month
        geo: global
        unit: 1
        price: 'call'
        userMultiplied: false

  - id: tensordock-instant-vms
    name: Instant VMs (beta)
    type: usage-based
    description: Pre-configured Instant VM templates with sub-30-second deployment.
    entries:
      - label: Instant VM hourly
        name: instant_vm_hour
        type: metered
        metric: vm_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: 'varies'
        userMultiplied: false

  - id: tensordock-containers
    name: Containers
    type: usage-based
    description: GPU container workloads deployable via the v0 container endpoints.
    entries:
      - label: Container GPU hourly
        name: container_gpu_hour
        type: metered
        metric: gpu_hour
        limit: -1
        timeFrame: hour
        geo: global
        unit: 1
        price: 'varies'
        userMultiplied: false
notes:
  - Billing is pro-rated to the microsecond.
  - When account balance hits $0, instances are automatically stopped and eventually deleted.
  - All prices in USD; published rates exclude applicable taxes (VAT, GST).
  - Marketplace dynamics mean the same GPU model can show different prices at different hosts.