Shared storage for your GPU instances.

Store datasets, model weights and checkpoints on a separate filesystem. Mount it on multiple instances in the same region, and keep your files when an instance is deleted.

$0.00014/GB/hour · Billed on provisioned capacity

Persistent storage

Download once. Reuse across instances.

Keep large files in one place as your compute needs change.

Reuse datasets

Attach the same filesystem to several of your instances. Each can read the datasets and model weights you have already stored.

Change GPUs

Launch a new instance with the filesystem attached. Your files stay available without another upload.

Keep checkpoints

Save results outside the instance disk. Pausing or deleting an instance leaves the filesystem intact.

Setup

Create it, then attach it.

Choose a region and capacity. Attach the filesystem when you launch or resume an instance.

CLI example
jl filesystem create --name datasets --storage 200

# Use the ID returned by the create command
jl create --gpu A100 --fs-id <fs_id>
Filesystem specifications
SettingHow it works
Size50 GB to 10 TB, grown in 50 GB steps
ResizingGrow anytime. Never shrinks
Mount path/home/jl_fs/ on every attached instance
AttachAt instance launch or resume, by fs-id
Concurrent mountsAs many of your own instances as you need
LifetimeIndependent: survives pause, resume and destroy
RegionLives in one region; attaches to instances in that region

Storage choices

Choose where your files live.

Use instance storage for your environment and scratch files. Use a filesystem for data you need across instances.

Storage lifecycle
Instance storageFilesystem
Attached toOne instanceMultiple instances in your account
When you pauseRetained; storage charges continueRetained; storage charges continue
When you delete an instanceDeleted with the instanceRetained until you delete the filesystem
RegionInstance regionMust match attached instances

In our ComfyUI test on H100, first model load took 1.71 seconds from instance storage and 1.73 seconds from a filesystem. This measures one workload, not general storage throughput. Read the test

Pricing

Pay for the capacity you provision.

Storage costs $0.00014 per GB per hour, whether full or empty. Charges continue when no instances are attached.

A 200 GB example

At 720 hours, a 200 GB filesystem costs $20.16. GPU instance charges are separate.

Grow when needed

Increase capacity in 50 GB steps, up to 10 TB. Filesystems cannot shrink.

Delete when finished

Deleting the filesystem stops its storage charges and permanently removes its files. Save anything you need first.

Before you start

Common questions.

Practical details for choosing and using this product.

Can I attach a filesystem to a running instance?

Attach it at launch or resume. For a running instance, pause it, select the filesystem, then resume.

Can I share it with another account?

Filesystems attach to your own instances in the same region. They are not shared across accounts or team members.

Do I need to change my code?

Files appear at /home/jl_fs/. Point your application at that path, or use symlinks. You can use ordinary file operations without a separate storage API.

What happens if I delete an instance?

The filesystem and its charges continue independently. Only deleting the filesystem removes those files and ends its billing.

Keep your data ready for the next run.

Create a filesystem in the region where you run your GPUs.