The current pricing page asks customers to request Sandbox access through sales. The CPU/GPU sandbox path and the separate persistent Instances product have different storage and lifecycle contracts. Read the source ↗

Agent sandboxes / Thunder Compute Sandboxes and Instances alternative

A Thunder Compute sandbox alternative for returning Linux agents

Thunder Compute now offers agent-oriented CPU/GPU sandboxes alongside persistent GPU Instances. Sandboxes run commands and transfer files, but their filesystem is destroyed when they end. Instances have a separate disk-snapshot workflow. Compare these product paths independently when choosing how an agent’s installed tools and files return.

Start buildingCompare capabilities ↓

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

When Thunder Compute Sandboxes and Instances fits

Thunder fits GPU-backed agent jobs, evaluations and development workloads that need its accelerator capacity. Its instance snapshots already retain disk state, while its sandbox SDK provides direct command execution and outbound network controls.

When to consider Sandbox0

Consider Sandbox0 for compatible Linux agents requiring a complete durable writable RootFS across tasks, enabled automatic online growth and supported file operations while compute stays paused. Retain suitable GPU execution separately.

Capabilities

Thunder Compute Sandboxes and Instances and Sandbox0: what persists?

Feature comparison with documentation and verification status
Environment and runtime lifetimeThunder Compute Sandboxes and InstancesNot 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 filesThunder Compute Sandboxes and InstancesThe separate Instances product retains disk through snapshots; sandbox files remain ephemeral and are destroyed when the sandbox ends.View evidence ↗Sandbox0The writable RootFS survives successful checkpoints, including software, system configuration, dependencies, and user data.Read documentation →
Whole RootFSThunder Compute Sandboxes and InstancesThe agent-sandbox path destroys its filesystem on termination; instance disk snapshots are a separate retained-OS workflow.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 stateThunder Compute Sandboxes and InstancesInstances use snapshot/delete/create rather than native Stop; sandbox stop/termination is not a documented durable 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 stateThunder Compute Sandboxes and InstancesInstance disk snapshots and sandbox command reconnection do not establish saved guest RAM or GPU-memory 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.

The sandbox SDK selects ephemeral storage size, while Instances retain selected disk snapshots. Neither path establishes automatic online growth of a complete durable sandbox RootFS 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.

Thunder’s sandbox upload/download interfaces operate on the live sandbox or image-backed container. Instance snapshots retain disk but do not establish direct file edits on a stopped instance root. Sandbox0 supports its paused durable RootFS under filesystem-only mutation and retained-memory read rules.

See paused file operations →

Sandbox expiry is configurable, but termination destroys the disk

The SDK defaults sandbox lifetime to 300 seconds measured from creation and permits disabling automatic expiry. It also explicitly says termination destroys the filesystem. Commands can survive client disconnects, which is useful continuity while the sandbox remains alive, but does not preserve its files after termination.

Source 1 ↗Source 2 ↗

Instance snapshots preserve disk through a replacement workflow

The Instances guide says there is no native Stop operation: create a disk snapshot, delete the instance and later create a new one from the snapshot. The snapshot guide captures persistent-disk state, and the retained snapshot continues billing until deleted. This is useful OS-disk continuity, with no established RAM/GPU-state continuation.

Source 1 ↗Source 2 ↗

Sandbox and instance rates use different resource bundles

The sandbox H100 rate is per GPU plus separate vCPU, GiB memory and GiB storage charges. The instance H100 price has its own included CPU/RAM and disk allowances. The current sandbox SDK supports one GPU, while instance multi-GPU options are separate. Sandbox access is requested through sales rather than assumed instant general availability.

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.

Thunder Compute Sandboxes and Instances

Pricing reviewed 2026-10-19

Sandbox H100 SXM
$2.95 per GPU-hour
GPU component only; CPU, memory and storage are additional.
Rate source ↗
Sandbox CPU and memory
$0.051/vCPU-hour plus $0.016/GiB-hour
Rate source ↗
Sandbox ephemeral storage
$0.00012 per GiB-hour
Rate source ↗
H100 SXM Instance reference
$3.20 per GPU-hour
Instance configuration and included resources differ from sandbox resources.
Rate source ↗
Retained instance snapshots
$0.05 per GB-month
Metered hourly; monthly reference assumes 730 hours.
Rate source ↗

Sandboxes are billed per second; Instances and extra instance resources are billed per minute while running. Snapshot storage continues separately until deletion. Preserve the vendor’s GB versus GiB units and selected instance allowances; do not compare a GPU-only sandbox rate with a complete configured machine.

Check current Thunder Compute Sandboxes and Instances 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 Thunder Compute Sandboxes and Instances

  1. Separate agent sandboxes from persistent Instances and record resource sizes, optional expiry, network policy and snapshot requirements.
  2. Export sandbox files before termination and retain required GPU services; use verified disk snapshots for instance migration where needed.
  3. Test durable system paths, online growth, paused files and recovery independently from GPU execution and product-specific resource rates.

Common questions

Is Sandbox0 a drop-in Thunder Compute Sandboxes and Instances replacement?

Evaluate API compatibility and product scope. A shared sandbox category does not make SDKs, lifecycle semantics, or limits interchangeable.

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 ↗www.thundercompute.com/docs/sandboxes/python-sdk.md ↗www.thundercompute.com/products/sandboxes ↗www.thundercompute.com/docs/guides/stopping-instances.md ↗www.thundercompute.com/docs/vscode/operations/snapshots ↗www.thundercompute.com/pricing ↗Sandbox0 RootFS capacity →

Related comparisons

Agent 37 vs Sandbox0 →AgentComputer vs Sandbox0 →ainclave vs Sandbox0 →Alibaba Cloud Agent Sandbox / FC / AgentRun vs Sandbox0 →Alibaba Cloud AgentBay vs Sandbox0 →Amika vs Sandbox0 →