# TensorDock

Official pricing: https://www.tensordock.com/  
Category: gpu-cloud · Isolation: vm

## Pricing regimes (raw)

- **H100 SXM5 on-demand (marketplace floor)** (resource): $0.003/vCPU-h, $0.002/GiB-h [stock]
- **A100 SXM4, VRAM unverified, on-demand** (resource)
- **RTX 4090 on-demand (marketplace floor)** (resource): $0.003/vCPU-h, $0.002/GiB-h [stock]
- **H100, spot** (resource) [spot]
- **A100, spot** (resource) [spot]
- **RTX4090, spot** (resource) [spot]
- **CPU Cloud (published floor)** (sizes) [stock]
- **Instant VMs** (sizes) [beta]

## Features

Yes: pause/resume, persistent disk, ≥24 h sessions, full VM (own kernel), Docker inside, root, desktop GUI, SSH, public IPv4, raw TCP inbound, open internet, static egress IP, GPU, Windows, Python, EU data residency, gVisor or VM (no shared kernel), inbound access rules, firewall inside (nftables)

No: snapshots, memory snapshots, fork/clone, volumes, idle auto-stop, start from your own snapshot, your own Docker/OCI image, browser, computer-use API, browser + desktop control, anti-bot stealth, CAPTCHA solving, residential IPs, preinstalled agents, HTTPS preview URLs, egress allowlist, self-hosting, BYOC, open source, SOC 2, HIPAA, SSO, arm64, macOS, Node.js, HTTP method/path egress rules, wake on request, live resize, live fork (no pause), memory fork, MCP server, automatic snapshots, snapshots on demand, agent harness API, hosted agent API (their own agent), secret proxy, secret proxy for any API, extra volumes, shared volumes

Unknown: everything else. Evidence (source + quote) per feature: https://battleships.dev/data/providers/tensordock.json → feature_evidence

## Caveats

- Full root OS control, dedicated passthrough GPU, Linux/Windows and Docker templates. GPU price is set per host (the published "From" price is the floor); CPU $0.003/vCPU-h, RAM $0.002/GB-h, storage $0.00005/GB-h ($0.0365/GB-month, also billed while stopped) and dedicated IPv4 $0.006/h ($4.38/month) are flat calculator constants on the deploy page. $5 deposit; zero balance automatically deletes servers.
- Any null count, CPU/RAM, billing increment or minimum is unverified, not unlimited/free.
- This is GPU-product research, not a claim that every feature of the entire vendor documentation was audited. Unverified features are null.
- GPU multi-count bundles, variant/region selection and fractional slices require a SKU-aware estimator; unsupported rows remain unpriced in strict cards.
- Modes with gpu=null (A100 with unverified VRAM, spot) stay unpriced, not free. Never substitute 0 for a null rate. On-demand GPU VMs are ordinary GPU instances (no alt flag); marketplace availability is shown via the stock flag.
- Unknown fee, storage or bandwidth means full monthly total may be unavailable. Rates exclude taxes.
- GPU rates are marketplace floors ("From $X/hr"): a specific host/region can cost more, and the cheapest host may be sold out. Estimates are a lower bound.

## How this provider charges

As of 2026-09-28. KVM GPU VMs, independent-host marketplace.

Full root OS control, dedicated passthrough GPU, Linux/Windows and Docker templates. Host sets GPU, CPU, RAM and disk prices. $5 deposit; zero balance automatically deletes servers.

| Regime | When applicable | Billing | Numbers ($/GPU-hour unless slice) | Source |
|---|---|---|---|---|
| on-demand  | H100 SXM5; count None; Region/availability must be checked at launch. | unknown; min unknowns | 2.25; list-from; per_gpu_hour | [source](https://www.tensordock.com/) |
| on-demand  | A100 SXM4, VRAM unverified; count None; Region/availability must be checked at launch. | unknown; min unknowns | 1.8; list-from; per_gpu_hour | [source](https://www.tensordock.com/) |
| on-demand  | RTX4090 ; count None; Region/availability must be checked at launch. | unknown; min unknowns | 0.35; list-from; per_gpu_hour | [source](https://www.tensordock.com/) |
| spot  | H100 ; count None; Region/availability must be checked at launch. | unknown; min unknowns | unknown; dynamic-bid; per_gpu_hour | [source](https://docs.tensordock.com/virtual-machines/spot-instances.md) |
| spot  | A100 ; count None; Region/availability must be checked at launch. | unknown; min unknowns | unknown; dynamic-bid; per_gpu_hour | [source](https://docs.tensordock.com/virtual-machines/spot-instances.md) |
| spot  | RTX4090 ; count None; Region/availability must be checked at launch. | unknown; min unknowns | unknown; dynamic-bid; per_gpu_hour | [source](https://docs.tensordock.com/virtual-machines/spot-instances.md) |

## Gotchas
- Full root OS control, dedicated passthrough GPU, Linux/Windows and Docker templates. Host sets GPU, CPU, RAM and disk prices. $5 deposit; zero balance automatically deletes servers.
- Any null count, CPU/RAM, billing increment or minimum is unverified, not unlimited/free.
- This is GPU-product research, not a claim that every feature of the entire vendor documentation was audited. Unverified features are null.
- GPU multi-count bundles, variant/region selection and fractional slices require a SKU-aware estimator; unsupported rows remain unpriced in strict cards.

## Worked example (required protocol workload)
4 vCPU / 8 GiB, 50 concurrent × 8 hours/day × 22 days = **8,800 instance-hours/month**. 30% CPU utilization does not reduce allocated GPU uptime. 50 GiB snapshots and 100 GiB egress are additional. The requested CPU-only workload has no GPU type/count, so its full GPU-provider total is **null**, not a fictitious CPU equivalent.

No fixed single-full-GPU USD quote suitable for a numeric example was verified. Compute = 8,800 × selected offer $/hour, or required node count × node price; keep the estimate null until the offer/FX/commitment is resolved.

## Evidence scope
Sources read are listed in the provider/features JSON. Raw pricing pages, relevant docs and API payloads are retained under `../raw/`. All unknown feature toggles remain null.