Development environments / Kaggle Notebooks alternative

Kaggle Notebooks alternative

Kaggle Notebooks combine hosted code, datasets, competitions and CPU/GPU/TPU execution. Saved notebook versions and working-directory outputs are useful durable artifacts. The current documentation explicitly distinguishes them from scratch space and a newly created clean execution session.

Start buildingCompare capabilities ↓

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

When Kaggle Notebooks fits

Kaggle fits notebook analysis, competition submissions, dataset pipelines and collaborative model experiments. Preserve its data/community workflow and accelerator access where those are central; Sandbox0 is a different Linux execution layer.

When to consider Sandbox0

Consider Sandbox0 when a returning agent needs installed system tools and configuration retained beyond a notebook’s saved output directory, with online RootFS growth and supported paused-file operations. You provide the notebook/UI and specialized compute integration separately.

Capabilities

Kaggle Notebooks and Sandbox0: what persists?

Feature comparison with documentation and verification status
Environment and runtime lifetimeKaggle NotebooksDocumented 12-hour CPU/GPU and 9-hour TPU sessions are bounded, not indefinitely running machines.View evidence ↗Sandbox0ttl 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 filesKaggle NotebooksSaved outputs up to 20 GB under /kaggle/working can be reused; scratch space outside it is session-scoped.View evidence ↗Sandbox0The writable RootFS survives successful checkpoints, including software, system configuration, dependencies, and user data.Read documentation →
Whole RootFSKaggle NotebooksThe docs explicitly exclude scratch space outside the saved working directory from cross-session 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 stateKaggle NotebooksNotebook versions and clean re-execution do not establish arbitrary-process pause/resume.View evidence ↗Sandbox0Pause compute between tasks; resume on available capacity. Default resume starts fresh processes from the saved filesystem.Read documentation →
Memory and process stateKaggle NotebooksQuick Save captures displayed notebook work, not an established full process-memory checkpoint.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.

Kaggle’s saved 20 GB working outputs and Docker-image-backed reruns are acknowledged. The reviewed guide does not establish automatic online expansion and retention of the entire writable RootFS. System tools outside the saved-output boundary need an explicit rebuild or another runtime.

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.

Saved versions and outputs can be accessed after execution, but that is an artifact workflow rather than operations on every path of a paused Linux RootFS. Sandbox0’s supported paused-file interface has a different scope.

See paused file operations →

The saved-output boundary is explicit

The documentation allows up to 20 GB of outputs under /kaggle/working to be saved and reused as notebook inputs. Extra scratch space outside that path is not saved beyond the session. That is useful persisted data with a directory boundary, rather than preservation of all system packages and every writable RootFS path.

Source 1 ↗

Quick Save and Save & Run All are different operations

Quick Save records the notebook as displayed. Save & Run All starts a separate clean session and executes top to bottom, producing a version associated with a Docker image. It does not continue every live Python object or arbitrary process from the interactive session.

Source 1 ↗

Execution and idle limits affect long agent work

The current technical specification lists 12-hour CPU/GPU and 9-hour TPU sessions, plus 20 minutes of interactive idle time when code is not modified or executed. Hardware can vary, and accelerator quotas apply. Save required work within those limits rather than assuming every long task can be kept alive indefinitely.

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.

Kaggle Notebooks

Pricing reviewed 2026-10-18

Documented P100 GPU option
Available free, subject to quotas
The official notebook guide describes a free single-P100 option; actual allocation may vary.
Rate source ↗
Documented TPU option
Available free, subject to quotas
Session, availability and account conditions still apply.
Rate source ↗
Connected cloud products
Separate provider charges
The guide warns that integrations such as Google Cloud products can require billing; notebook access does not make them free.
Rate source ↗

Listed hardware is a documentation specification, not guaranteed dedicated capacity or throughput. Promotional bonus accelerator hours depend on eligibility and are not treated as a universal allowance. No numerical compute-hour price or GPU equivalence with Sandbox0 is invented.

Check current Kaggle Notebooks 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 Kaggle Notebooks

  1. Export notebooks, required /kaggle/working outputs and external data references; inventory scratch files, installed tools and accelerator requirements.
  2. Rebuild compatible tools in a Sandbox0 template and integrate notebook/data/accelerator services explicitly.
  3. Verify repeated-task state, whole-RootFS retention and growth, paused files, access permissions and complete compute/storage costs.

Common questions

Is Sandbox0 a drop-in Kaggle Notebooks 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.kaggle.com/docs/notebooks ↗Sandbox0 RootFS capacity →

Related comparisons

Agency Tool Company vs Sandbox0 →Amp Orbs vs Sandbox0 →Bitrise RDE vs Sandbox0 →boxes.dev vs Sandbox0 →Capy vs Sandbox0 →Claw 2 Agent vs Sandbox0 →