Your GPUs.
Run open-weight models on your own infrastructure. Connect vLLM, SGLang, or another OpenAI-compatible server and make the hours between jobs count.
There’s potential in the hours between jobs. Put your available inference capacity on the exchange and connect with builders who need it.
Become a provider ↗Run open-weight models on your own infrastructure. Connect vLLM, SGLang, or another OpenAI-compatible server and make the hours between jobs count.
Bring your inference service to a new audience. List the capacity you operate, control your pricing, and let eligible buyer requests find you.
Make use of reserved throughput you have permission to resell. Turn available capacity into a listed offer, with clear limits and availability.
Add your endpoint and dedicated credential. Discover the models you can serve and match them to the catalog.
Choose your price, schedule, and concurrency cap. Use auto-pricing with a floor you control.
Your offer runs through behavioral verification before listing. The exchange credits verification traffic.
Receive eligible requests and earn on completed calls. Track earnings and request a payout from $5.
Pause an offer, rotate a credential, or delist from your dashboard.
One published 5% exchange fee. No hidden spread.
Prices stay at or below the public list price when known.
Auto-pricing adjusts within your limits as competition and utilisation change. Your floor is always respected.
Define selling windows around your own workloads. Add concurrency and daily caps so new demand fits your operation.
Publish prepaid price locks for a defined block of capacity. Buyers get fixed rates, and you get a clearer view of committed demand.
An OpenAI-compatible endpoint, a dedicated credential, and the right to sell the listed capacity. Payout requests are currently processed manually. See the provider guide for setup details.