Create Compute Instance
Creates a new compute instance with the specified configuration and SSH key.
Requirements:
- Requires compute permissions (extra_permissions.compute = true)
- Authentication required via admin API key
- Valid SSH public key required for instance access
Key Features:
- Create high-performance GPU instances
- Specify sector for InfiniBand configuration (8x H100 only)
- SSH key-based authentication
- Automatic instance provisioning and region assignment
- Idempotent creation with Idempotency-Key header (optional but recommended)
Common Use Cases:
- Spin up compute resources for ML training
- Create GPU instances for inference workloads
- Set up development environments with H100 GPUs
- Deploy distributed training with InfiniBand networking
Instance Types:
gpu_8x_h100_sxm5: 8x NVIDIA H100 GPUs (high-performance, supports sector configuration for InfiniBand)gpu_1x_h100_sxm5: 1x NVIDIA H100 GPU (standard)
Idempotency:
- Optional Idempotency-Key header prevents duplicate instance creation on retries
- Responses cached for 10 minutes per unique key
See fal.ai docs for more details.
Documentation Index
Fetch the complete documentation index at: https://fal.ai/docs/llms.txt
Use this file to discover all available pages before exploring further.
Authorizations
Admin API key must be prefixed with "Key ", e.g. Authorization: Key YOUR_ADMIN_API_KEY
Headers
Optional idempotency key for safe request retries
"550e8400-e29b-41d4-a716-446655440000"
Body
Request body for creating a new compute instance with SSH access
Type of compute instance to create
gpu_8x_h100_sxm5, gpu_1x_h100_sxm5 "gpu_1x_h100_sxm5"
SSH public key for accessing the instance (e.g., 'ssh-rsa AAAAB3...')
1"ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABgQC... user@host"
Sector for InfiniBand configuration (only valid with gpu_8x_h100_sxm5)
sector_1, sector_2, sector_3 "sector_1"
Response
Compute instance created successfully
Compute instance details including configuration, location, and status
Unique identifier for the compute instance
"inst_abc123xyz"
Type of compute instance (GPU configuration)
gpu_8x_h100_sxm5, gpu_1x_h100_sxm5 "gpu_1x_h100_sxm5"
Geographical region where the instance is located
us-west, us-central, us-east, eu-north, eu-west, other "us-west"
Current operational status of the instance
ready, init, pending, provisioning, stopped, unknown "ready"
Sector identifier for instance placement within the region (if applicable)
sector_1, sector_2, sector_3 "sector_1"
IP address of the instance (available when instance is ready)
"203.0.113.42"
Nickname of the user who created this instance
"developer"