GPU clouds / Fluidstack alternative

A Fluidstack alternative for durable agent orchestration

Fluidstack builds and operates large-scale AI compute infrastructure. Its documentation describes managed bare-metal Kubernetes and Slurm for training and inference. That cluster service is adjacent to the per-user Linux environment an agent returns to between tasks; assess both layers independently.

Start buildingCompare capabilities ↓

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

When Fluidstack fits

Fluidstack fits teams procuring or operating substantial accelerator capacity and managed cluster infrastructure, with specialized engineering support for frontier training and inference.

When to consider Sandbox0

Consider Sandbox0 for compatible Linux agents needing a complete durable writable RootFS, automatic online growth, and supported paused-file access. Retain accelerator capacity and specialized cluster operations for the workloads that need them.

Capabilities

Fluidstack and Sandbox0: what persists?

Feature comparison with documentation and verification status
Environment and runtime lifetimeFluidstackNot 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 filesFluidstackDurability depends on the selected cluster storage and agreement; a current universal disk-retention guarantee is not established.View evidence ↗Sandbox0The writable RootFS survives successful checkpoints, including software, system configuration, dependencies, and user data.Read documentation →
Whole RootFSFluidstackLarge-scale infrastructure operation does not establish complete per-user writable RootFS retention.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 stateFluidstackVerify actual cluster and workload restart semantics; no uniform compute-free agent pause contract is established.View evidence ↗Sandbox0Pause compute between tasks; resume on available capacity. Default resume starts fresh processes from the saved filesystem.Read documentation →
Memory and process stateFluidstackManaged cluster operation does not itself establish durable GPU or process-memory checkpoints.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.

Storage expansion in a Fluidstack cluster depends on the contracted storage and operating model. That is a different contract from automatic online expansion of each Sandbox0 agent’s enabled writable RootFS.

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.

Verify data-access methods for the actual cluster storage separately from whether a workload is running. Sandbox0 provides supported paused operations on its durable writable RootFS, rather than requiring the user to infer file access from a cluster stop command.

See paused file operations →

Managed GPU infrastructure is the primary scope

Current company materials emphasize building and operating AI compute at scale. The documentation describes Kubernetes and Slurm on bare metal, GPU-related operators, health checks and observability. Those services address cluster reliability and workload placement rather than automatically specifying a private persistent agent computer.

Source 1 ↗Source 2 ↗

Obtain the applicable storage and restart contract

Cluster storage, node-local disks, job checkpoints and application backups have different failure and retention boundaries. Confirm the actual filesystem, backup policy, node replacement behavior and supported restart operations in the selected agreement. Managed cluster operations and job restarts do not by themselves establish saved GPU memory or a complete offline RootFS interface.

Source 1 ↗

A configured capacity quote is required

The current public homepage does not publish a numeric GPU-hour rate card. Obtain terms for GPU model and count, topology, storage, support, deployment location and commitment. Do not compare old marketplace rates or a spot quote with a current managed-capacity agreement.

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.

Fluidstack

Pricing reviewed 2026-10-19

Configured GPU or managed-cluster capacity
Obtain a current agreement-specific quote
GPU count, commitment, storage, support and topology determine the scope.
Rate source ↗
Stopped or reserved capacity
Confirm effective billing terms
A stop operation should not be assumed to cancel a capacity commitment.
Rate source ↗

No current public numeric rate card was verified. Compare complete agreed infrastructure scope and retained capacity, rather than a nominal GPU device-hour with an unrelated CPU-only agent shape.

Check current Fluidstack 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 Fluidstack

  1. Inventory the capacity agreement, GPU jobs, cluster schedulers, durable storage, local scratch, identities and observability requirements.
  2. Keep GPU-heavy workloads on suitable infrastructure and port compatible Linux agent orchestration with verified exports of its own system and files.
  3. Test long tasks, restart boundaries, growing state, supported paused files and the combined GPU, agent and retained-storage budget.

Common questions

Is Sandbox0 a drop-in Fluidstack 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 ↗fluidstack.io/ ↗docs.fluidstack.io/ ↗fluidstack.io/news/fluidstack-awarded-clustermax-gold-by-semianalysis ↗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 →Google Compute Engine GPU VMs vs Sandbox0 →Hetzner GPU servers vs Sandbox0 →