Permanent storage for AI datasets and research
Datasets, model checkpoints, research outputs. Data that took months to produce deserves storage that does not expire.
Store Research DataDatasets stay available
Keep benchmark sets, corpora and research data accessible for years — reproducibility depends on it.
Verifiable provenance
Transaction IDs and hashes let reviewers verify exactly which dataset version was used.
Citation-friendly
Each dataset gets a permanent URL you can cite in papers and model cards.
01The reproducibility problem
Research findings are only as durable as their underlying data. When datasets live on personal drives or temporary links, papers become impossible to reproduce.
Long Drive gives datasets a permanent home on Arweave — with a timestamped, immutable record of what was stored and when.
02What researchers store
Small but critical artifacts: evaluation sets, prompt templates, configuration files, metadata. Files under 100 KB are free forever — ideal for configs and small corpora.
Large datasets incur a one-time fee based on current rates. For non-public data, enable browser encryption before upload.
03The practical case
A research lab archived its evaluation harness — configs, prompt sets and small benchmark fixtures, most under 100 KB. Total cost: zero. Every artifact now has a permanent, citable URL.
Reviewers can fetch the exact versions referenced in the paper's appendix, verified against the Arweave ledger.
AI & research FAQ
Can I store a large model checkpoint?
Yes, with a one-time fee based on size — check the calculator first. Checkpoints are often hundreds of MB to GBs.
Is research data public?
Unless you encrypt it, yes — which suits open datasets. Use browser encryption for private or embargoed data.
How do I cite a dataset?
Cite the permanent URL along with the transaction ID — both are stable and verifiable.