Cerebrium CPU/GPU memory checkpointing is early beta. Its guide warns of capture/restore failures; ephemeral filesystem files are not included in the checkpoint. Read the source ↗

GPU clouds / Cerebrium Serverless Compute alternative

Cerebrium serverless alternative for returning agent state

Cerebrium runs serverless compute with project-level persistent volumes and early-beta CPU/GPU memory checkpointing. Those checkpoints can restart from saved execution state, but explicitly do not copy ephemeral files. Separate the model/container startup optimization from a returning agent’s complete durable system.

Start buildingCompare capabilities ↓

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

When Cerebrium Serverless Compute fits

Choose Cerebrium for serverless model/compute workloads, accelerator options and selected persistent data volumes. Its memory checkpoint beta is useful to evaluate for expensive model initialization, with the documented restore and compatibility limits.

When to consider Sandbox0

Consider Sandbox0 for compatible Linux agents whose installed tools, system configuration and outputs should live on one durable growing RootFS with supported paused-file access. Keep GPU/model stages on suitable compute and account for both services.

Capabilities

Cerebrium Serverless Compute and Sandbox0: what persists?

Feature comparison with documentation and verification status
Environment and runtime lifetimeCerebrium Serverless ComputeNot 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 filesCerebrium Serverless ComputeProject volumes preserve selected files across deployments and support external CLI file operations.View evidence ↗Sandbox0The writable RootFS survives successful checkpoints, including software, system configuration, dependencies, and user data.Read documentation →
Whole RootFSCerebrium Serverless ComputeMemory checkpoints do not copy ephemeral filesystem writes; durable files use the separate mounted storage scope.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 stateCerebrium Serverless ComputeDeployment-specific checkpoint restoration is not a documented general returning-machine pause/resume contract.View evidence ↗Sandbox0Pause compute between tasks; resume on available capacity. Default resume starts fresh processes from the saved filesystem.Read documentation →
Memory and process stateCerebrium Serverless ComputeEarly-beta checkpointing captures CPU and GPU memory; ephemeral files and TCP connections are excluded and restores may fail.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.

Cerebrium already provides project volumes with explicit self-service enlargement up to 1 TB. That selected mount and a memory-only checkpoint are distinct from automatic online expansion of an entire durable 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.

Cerebrium already provides external file upload/download/list operations for project storage. Their scope is the mounted volume, not arbitrary files throughout a paused complete OS; ephemeral files are excluded from memory checkpoint restoration.

See paused file operations →

Project volumes preserve selected mounted files

Cerebrium mounts project storage at /persistent-storage and provides CLI upload/download/list operations. Normal volumes can be resized through self-service up to 1 TB; global-volume behavior differs. This is persistent selected storage, rather than automatic coverage of every mutable OS path.

Source 1 ↗

The checkpoint beta captures memory, not ephemeral files

CPU/GPU memory is saved and execution resumes from the checkpoint request point. Files written to the ephemeral filesystem are not copied, TCP connections disconnect and the checkpoint is tied to a deployment. GPU dumps also require enough container RAM beyond normal application usage.

Source 1 ↗

A complete quote includes several resource meters

GPU, vCPU and memory have separate usage rates; storage and an optional paid platform tier add other meters. The pricing page illustrates billing allocated resources while requests are processed. Confirm dedicated/guaranteed-capacity commitments and the workload’s effective lifecycle before estimating the total.

Source 1 ↗

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.

Cerebrium Serverless Compute

Pricing reviewed 2026-10-19

H100 GPU reference
$0.000944 per GPU-second
GPU component; add selected CPU/memory resources.
Rate source ↗
CPU resource reference
$0.00000655 per vCPU-second
Rate source ↗
Memory resource reference
$0.00000222 per GB-second
GB unit from the price table; checkpoint GPU dumps can require more host memory.
Rate source ↗
Persistent storage
$0.05 per GB-month; first 100 GB free
Confirm effective volume allocation and any special storage terms.
Rate source ↗
Standard platform tier
$100/month + compute
Hobby has no platform subscription fee but still charges compute; plan entitlements differ.
Rate source ↗

Do not treat GPU-only pricing as the full workload bill or convert GB to GiB silently. Include CPU, memory, volume storage, platform tier and any capacity commitment. Checkpointing is early beta with possible capture/restore failures and deployment coupling.

Check current Cerebrium Serverless Compute 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 Cerebrium Serverless Compute

  1. Inventory deployed app/resources, project/global volumes, files outside the mount, checkpoint deployment coupling and CPU/GPU/memory costs.
  2. Keep needed GPU/model stages and move compatible Linux agent orchestration plus required durable files to the appropriate runtime.
  3. Test checkpoint restore limitations, system retention, growth, paused files and full compute/memory/storage/platform plus Linux runtime costs.

Common questions

Is Sandbox0 a drop-in Cerebrium Serverless Compute 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 ↗cerebrium.ai/docs/storage/managing-files.md ↗cerebrium.ai/docs/performance/checkpointing.md ↗cerebrium.ai/pricing ↗Sandbox0 RootFS capacity →

Related comparisons

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