GPU clouds / TensorDock GPU VMs alternative

A TensorDock alternative for the agent around GPU work

TensorDock offers root-access KVM VMs with dedicated GPUs and a marketplace of independently priced hosts. Its Core Compute locations use network storage so a stopped VM can return on another compatible server in the same location. This already provides meaningful OS-disk continuity, with a different operating model from a managed persistent agent runtime.

Start buildingCompare capabilities ↓

Reviewed 2026-10-19 · Published by the Sandbox0 team

When TensorDock GPU VMs fits

TensorDock fits GPU workloads needing a full VM, driver control and choices of host, location and accelerator. Select Core Compute network storage when same-location disk continuity across compatible hosts is required; other marketplace VMs have different placement constraints.

When to consider Sandbox0

Consider Sandbox0 for compatible Linux agents around GPU work, with a complete durable writable RootFS, automatic free-space-triggered growth and supported RootFS files while guest compute stays paused. Keep suitable GPU capacity separately.

Capabilities

TensorDock GPU VMs and Sandbox0: what persists?

Feature comparison with documentation and verification status
Environment and runtime lifetimeTensorDock GPU VMsNot established in this reviewSandbox0ttl and hard_ttl default to 0 (disabled). Configure idle pause and deletion policies for the same persistent sandbox identity; an idle running sandbox still accrues compute charges.Read documentation →
Persistent filesTensorDock GPU VMsCore Compute network storage retains the stopped VM and allows placement on another compatible server in the same location.View evidence ↗Sandbox0The writable RootFS survives successful checkpoints, including software, system configuration, dependencies, and user data.Read documentation →
Whole RootFSTensorDock GPU VMsThe retained Core Compute VM disk holds the selected OS and installed software; separate runtime mounts and provider-local storage need their own policy.View evidence ↗Sandbox0Persistence extends beyond /workspace to paths such as /usr/local, /etc, and /home. Runtime tmpfs and virtual filesystems are excluded.Read documentation →
Release compute, keep stateTensorDock GPU VMsCore Compute documents stopped VMs returning on compatible capacity; this is disk continuity rather than a RAM checkpoint guarantee.View evidence ↗Sandbox0Pause compute between tasks; resume on available capacity. Default resume starts fresh processes from the saved filesystem.Read documentation →
Memory and process stateTensorDock GPU VMsThe network-disk stop/start description does not establish durable RAM or GPU-memory snapshot restoration.View evidence ↗Sandbox0Experimental memory checkpoints can preserve compatible process state. Use memory: true on both pause and resume on a compatible deployment. A filesystem checkpoint alone does not preserve running processes.Read documentation →

Online RootFS growth

On growth-enabled images, Sandbox0 automatically expands the writable RootFS as free space runs low, within a configured sparse logical ceiling up to 64 TiB. Agent processes keep running. Growth is enabled in the qualified Cloud region; confirm activation for other regions. Existing images and snapshots retain their captured storage geometry. The ceiling is logical capacity, not preallocated host disk.

TensorDock exposes selectable VM CPU, memory and storage. The reviewed Core Compute contract does not establish automatic online growth of the whole boot filesystem as free space runs low.

Files while compute is paused

Sandbox0's Files API reads persistent files without resuming compute. Filesystem-only pauses also permit edits; retained-memory pauses are read-only and reject mutations with 409. SSH, SFTP, file watching, and runtime-only mounts require running compute.

See paused file operations →

Core Compute can retain the VM across hosts in one location

The Core Compute guide says network storage allows a stopped VM to deploy onto another server in the same location when that GPU is available. Distributed Compute is described as tied to its original server. Select the actual Core Compute label and verify network storage rather than applying one storage policy to every marketplace VM.

Source 1 ↗

A low advertised GPU price is not one universal configuration

The homepage lists starting rates for H100 SXM5, A100 SXM4 and RTX 4090. Hosts set their own prices, with location and redundancy differences. Compare the complete selected CPU, memory, GPU and storage configuration, not an advertised floor as guaranteed capacity.

Source 1 ↗

Spot interruption still leaves standard storage charges

The Spot guide describes beta interruptible instances with bids per VM, not per GPU. Storage stays at the host’s standard rate when compute is outbid. The homepage says a depleted prepaid balance leads to automatic server deletion, so retained state also depends on keeping the account funded.

Source 1 ↗Source 2 ↗

Pricing

Compare billing models, then price your workload

CPU allocation, memory, billing minimums, disk, plan fees, and supported workload differ. These meters are not equal-performance machine quotes. Confirm region, resource units, and included allowances before estimating a total.

TensorDock GPU VMs

Pricing reviewed 2026-10-19

H100 SXM5 advertised reference
From $2.25 per GPU-hour
Homepage floor; host, location and full configuration can differ.
Rate source ↗
A100 SXM4 advertised reference
From $1.80 per GPU-hour
Rate source ↗
RTX 4090 advertised reference
From $0.35 per GPU-hour
Rate source ↗
Spot retained storage
Host’s standard storage rate, including while outbid
The guide’s numeric examples illustrate bidding and are not a live storage quote.
Rate source ↗

Use the selected host’s actual running and stopped charges. Core Compute network storage, Distributed Compute placement and beta Spot interruption have different boundaries. A $0 prepaid balance can cause server deletion; advertised starting rates are not guaranteed inventory.

Check current TensorDock GPU VMs pricing ↗

Sandbox0

$0.015/GiB-hour compute (billed by the second at approximately $0.000004167/GiB-second), plus $0.02/GiB-month of retained incremental RootFS and memory checkpoint data. Shared base images and platform recovery retention are excluded; paused sandboxes accrue storage charges only.

Compute / GiB-hour
$0.015
Retained data / GiB-month
$0.02
Paused compute
$0.00

A 4 GiB agent, 100 running hours

$6.00 compute

4 GiB × 100 hours, billed by the second. Keeping 20 GiB of billable incremental data for a full billing month adds $0.40. CPU is derived from memory by the platform; choose a tier that fits your workload.

Understand storage billing →

Before migrating from TensorDock GPU VMs

  1. Record Core versus Distributed Compute, host/location, GPU shape, boot storage, Spot policy and account balance behavior.
  2. Export required tools and files and retain GPU execution where needed; verify whole-OS disk continuity under the selected VM policy.
  3. Port compatible Linux agents and test RootFS growth, supported paused files and recovery separately from GPU performance and host-specific billing.

Common questions

Is Sandbox0 a drop-in TensorDock GPU VMs replacement?

Evaluate API compatibility and product scope. These products serve different core workflows, so a change may require a different architecture.

Does keeping files also keep a running agent?

Filesystem state and volatile process state are separate. Sandbox0 requires explicit compatible memory checkpoints to preserve running processes; its default filesystem resume starts fresh processes.

Sources and review scope

The Boat team's Battleships directory is the discovery source. References below describe the product and plan discussed on this page. Check current availability, regional terms and billing units before migrating. Where the cited material does not establish a capability or price, the comparison identifies that uncertainty.

Battleships research record ↗docs.tensordock.com/virtual-machines/introduction-to-core-compute-vms.md ↗www.tensordock.com/ ↗docs.tensordock.com/virtual-machines/spot-instances.md ↗Sandbox0 RootFS capacity →

Related comparisons

Azure GPU VMs (NC_A100_v4 reference) vs Sandbox0 →Cerebrium Serverless Compute vs Sandbox0 →Crusoe vs Sandbox0 →DigitalOcean GPU Droplets vs Sandbox0 →Fluidstack vs Sandbox0 →Google Compute Engine GPU VMs vs Sandbox0 →