Volumes

The Volumes API lets you create and manage persistent storage attached to your GPU deployments. Volumes persist independently from deployments — you can delete a GPU instance and keep your data volumes to reuse with future deployments.

Base URL: https://gpu.huddleapis.com/api/v1

Volume Types

Type	Description	Use Case
NVMe	High-performance local NVMe SSD	Training datasets, model checkpoints, fast scratch storage
NVMe_Shared	Shared NVMe accessible across nodes	Multi-node training with shared data
HDD	Standard hard disk storage	Large datasets, archives, backups
HDD_Shared	Shared HDD across nodes	Shared bulk storage for multi-node setups
NVMe_Local_Storage	Instance-local fast storage	Temporary high-speed scratch space
NVMe_Shared_Cluster	Cluster-wide NVMe	Shared fast storage across an entire cluster
NVMe_OS_Cluster	OS disk for cluster deployments	Boot volumes for cluster instances

Create a Volume

curl -X POST -H "X-API-Key: mk_your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "training-data",
    "size": 500,
    "type": "NVMe",
    "location": "FIN-01"
  }' \
  https://gpu.huddleapis.com/api/v1/volumes

Response:

{
  "code": "OK",
  "data": {
    "id": "vol_abc123",
    "name": "training-data",
    "size": 500,
    "type": "NVMe",
    "location": "FIN-01",
    "status": "ordered",
    "is_os_volume": false,
    "contract": "hourly",
    "base_hourly_cost": 0.05,
    "currency": "USD",
    "created_at": "2026-04-03T10:00:00Z"
  }
}

Volume Lifecycle

Status	Description
`ordered`	Volume creation requested
`attaching`	Volume is being attached to a deployment
`attached`	Volume is mounted and accessible on the instance
`detached`	Volume exists but is not attached to any deployment
`deleted`	Volume has been permanently removed

Attaching Volumes to Deployments

You can attach volumes when creating a deployment:

New volumes created during deployment

{
  "cluster_type": "1H100.80S.30V",
  "hostname": "my-gpu",
  "image": "ubuntu-22.04-cuda-12.4-cluster",
  "ssh_key_ids": ["key-id"],
  "location": "FIN-01",
  "os_volume": {
    "name": "boot",
    "size": 100
  },
  "volumes": [
    {
      "name": "data",
      "size": 500,
      "type": "NVMe"
    }
  ]
}

Reuse existing volumes from a previous deployment

{
  "cluster_type": "1H100.80S.30V",
  "hostname": "my-gpu",
  "image": "ubuntu-22.04-cuda-12.4-cluster",
  "ssh_key_ids": ["key-id"],
  "location": "FIN-01",
  "existing_volumes": ["vol_abc123"]
}

Volume billing

Volumes are billed separately from GPU deployments. Storage costs accrue for as long as the volume exists, even when detached. Delete volumes you no longer need to stop charges.

List Volumes

# List all volumes
curl -H "X-API-Key: mk_your_key" \
  https://gpu.huddleapis.com/api/v1/volumes

# Filter by status
curl -H "X-API-Key: mk_your_key" \
  "https://gpu.huddleapis.com/api/v1/volumes?status=detached"

List Volume Types

To see which volume types are supported in a region:

curl -H "X-API-Key: mk_your_key" \
  https://gpu.huddleapis.com/api/v1/volume-types

Next Steps

Learn how to attach volumes in the Deployments guide
Check Regions to see volume type availability per region
Review GPU Billing for storage pricing details

Volumes

On this page