Every checkpoint travels with you
Training data stays in your account. Weights are yours to export and run wherever policy requires.
Weights are yours to export
Any checkpoint, any time, through the API, as a LoRA adapter, merged safetensors or GGUF.
Your data trains nobody else's model
Training datasets, checkpoints and finished weights stay inside your account.
Residency is a property of the project
Set a data region when you create a project and every file, run and checkpoint in it stays there.
Storage on one flat meter
Datasets, checkpoints and exported weights cost $0.10 per GB a month, metered by the GB-hour.
What you keep, entry by entry
Export converts a checkpoint into a LoRA adapter, merged safetensors or GGUF. Deliver it to a signed URL or to a Hugging Face repository you name. The export runs asynchronously and answers 202 with an operation you poll until it completes.
POST /v1/checkpoints/ckpt_000675/export
{ "format": "merged-safetensors", "destination": "signed-url" }
202 Accepted
{ "operation": "op_91c4", "format": "merged-safetensors", "status": "running" }downtime prevented this month · Predictive maintenance
monthly revenue misattributed · Ad performance
from site reading to client report · Field data
What the Own column includes
Export
- LoRA adapter
- Merged safetensors
- GGUF quantised format
- Signed URL delivery
- Hugging Face Hub delivery
- Raw archive download URL
Checkpoints
- Training and sampler types
- nucleus:// paths
- Expiry per checkpoint, null keeps
- Resolve by path
- Publish and unpublish, admin scope
- Permanent delete
Residency and security
- Region pinned per project
- Singapore (ap-southeast-1)
- AES-256 at rest, TLS in transit
- PDPA (Singapore)
- Scoped API keys
- Keys bound to a project
Where it runs
- Nucleus cloud
- Your cloud account, under licence
- On your premises, air-gapped
- One storage meter, $0.10 per GB / month
Storage is the meter this column runs on
Storage
Datasets, checkpoints and exported weights. Metered by the GB-hour, so you only pay while you keep them.
Five entries under Own
Weight export
Any checkpoint, any format
Checkpoints
Kept until you delete them
Data residency
Choose your region
Bring your own cloud
The platform in your account
Storage
Metered by the GB-hour
Read on
Docs: Checkpoints
Save, inspect, export and manage the weights produced by training runs.
Docs: Projects
Partition files, runs and checkpoints by team, and pin each partition to a data region.
Security at Nucleus
Nucleus is trusted with sensitive training data and private model weights by organisations of all sizes.
Keep what you train
Export any checkpoint and run it anywhere. Training data stays in your account.