Skip to main content
DataVendor is HUD’s marketplace for private AI training and evaluation data. Vendors list assets they already hold: codebases, tasksets with the RL environments they run on, and Raw data. Buyers, mostly AI labs, buy those listings, or post a request when the catalog does not have what they need. Every sale is between organizations, and vendors never learn which buyer bought. Vendors set their own prices, buyers see quality evidence before they pay, and access to a purchase lasts after the sale. How DataVendor works explains the model and the rules behind it.

Sell on DataVendor

Set up a vendor organization, put your supply in inventory, list it, and get paid.

Buy on DataVendor

Join as a buyer, evaluate listings, buy or sample them, or request data that does not exist yet.

​
Resources

Using the API

Team API keys, the buyer and vendor REST APIs, and the MCP server.

Glossary

Listings, assets, tiers, entitlements, and the rest of the vocabulary.

FAQ

What DataVendor is, how it makes money, and how it protects both sides.
DataVendor runs on the HUD platform. Tasksets and environments you sell or buy are the same objects the platform runs evaluations against, and a HUD team API key works with the DataVendor APIs too.