Agent testing / Refresh alternative

Refresh.dev alternative

Refresh builds artificial software worlds for evaluating and training agents, including computer-use applications, coding tasks and 3D workflows. Its environments and verifiable rewards help improve models. That is different from retaining a production agent’s complete Linux installation between user tasks.

Start buildingCompare capabilities ↓

Reviewed 2026-10-18 · Published by the Sandbox0 team

When Refresh fits

Refresh fits AI labs and teams seeking benchmark environments, tailored simulations or verified training data. Keep those artifacts and evaluation methods if the goal is measuring or post-training model capability.

When to consider Sandbox0

Consider Sandbox0 when you need a compatible execution layer with persistent system configuration, installed dependencies and growing whole RootFS. It can run suitable agent/harness code, but does not itself supply Refresh’s benchmark tasks, domain simulator or reward logic.

Capabilities

Refresh and Sandbox0: what persists?

Feature comparison with documentation and verification status
Environment and runtime lifetimeRefreshNot 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 filesRefreshPrepared benchmark environments and data do not establish a general-purpose durable disk contract.View evidence ↗Sandbox0The writable RootFS survives successful checkpoints, including software, system configuration, dependencies, and user data.Read documentation →
Whole RootFSRefreshWhole writable Linux RootFS persistence across arbitrary production tasks is not established.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 stateRefreshEvaluation/training environment reproducibility is not a documented arbitrary-machine checkpoint service.View evidence ↗Sandbox0Pause compute between tasks; resume on available capacity. Default resume starts fresh processes from the saved filesystem.Read documentation →
Memory and process stateRefreshArbitrary process-memory capture and restore are not established.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.

Refresh supplies task worlds and data for evaluation/training. The reviewed offer does not establish automatic expansion or complete retention of a production agent’s Linux RootFS. Sandbox0 addresses persistent system tools and growing files around compatible workloads.

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.

Benchmark/data delivery is different from accessing a paused Linux filesystem. The reviewed Refresh materials do not establish the complete paused-RootFS read/write contract; store required outputs under an explicit runtime or data agreement.

See paused file operations →

Different software worlds test different capabilities

Refresh’s homepage separates computer-use suites, Blender-style animation/simulation and coding work. Its overview also distinguishes terminal and computer-use formats. A runtime choice must satisfy the actual display, graphics and application requirements; the ability to execute generic Linux code does not establish compatibility with every simulator.

Source 1 ↗Source 2 ↗

A verifier and a retained production environment serve different goals

The overview describes task-specific verifiers and prepared developer workflows with resolved dependencies and containers. These support reproducible evaluation or training. They do not establish that every run returns to one customer-owned, growing Linux filesystem instead of the reset policy a benchmark needs.

Source 1 ↗

Bespoke environments and datasets need separate commercial scope

Refresh offers domain-specific worlds built on demand and invites labs to discuss environments and datasets. The reviewed public overview does not publish a generic CPU/RAM or storage tariff. Preserve task/data rights, verifiers, reset behavior and deployment obligations when estimating a complete system.

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.

Refresh

Pricing reviewed 2026-10-18

Environments and datasets
No public numeric usage tariff verified
Contact the vendor for the required environment, task/data scope, deployment and support terms.
Rate source ↗

The reviewed materials explain managed/bespoke simulation worlds and training data, rather than a self-service Linux resource price. Do not treat an available benchmark or prepared container as free managed infrastructure. Graphics, applications, training compute and licensed content can be separate costs.

Check current Refresh 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 Refresh

  1. Inventory the environment/application stack, task/verifier versions, reset rules, data rights and bespoke-service requirements.
  2. Run compatible agent/harness components on Sandbox0 while retaining the licensed simulator and any graphics or specialized training infrastructure separately.
  3. Check actual task outcomes and resets, retained system dependencies, RootFS growth, paused files and full data/application/compute costs.

Common questions

Is Sandbox0 a drop-in Refresh 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 ↗www.refresh.dev/ ↗refresh.dev/one-pager ↗www.refresh.dev/llms.txt ↗Sandbox0 RootFS capacity →

Related comparisons

Archal vs Sandbox0 →Arga Labs vs Sandbox0 →AutoComputer Research / Computer-use Services vs Sandbox0 →Chronicle Labs vs Sandbox0 →Halluminate vs Sandbox0 →HUD vs Sandbox0 →