Marketplace
Serve open models for pay
Run an open model on your own hardware, name your price, and the gateway sends you requests when you are the better buy. You are paid what your price says for every token you served.
How it works
- 1Apply with your endpoint, the models it serves, your price per 1M tokens in and out, and the wallet you want to be paid in. The operator reviews every application.
- 2Before any customer request reaches you, the gateway measures your endpoint: how fast it answers, whether the tool calling and JSON output you declare work, a few task probes, answers to fixed questions compared with a reference model, and whether the token counts you report match the gateway's own count. A failing endpoint is paused with the reason and checked again later.
- 3A verified endpoint competes with the house route on price, measured speed and measured quality. Customers pay your price plus the gateway's margin, and never learn which provider served them.
- 4Each request you serve adds what your price says for its tokens to what you are owed. The operator sends payouts by hand to your wallet and records each transaction here.
- 5Say what you keep. Until the operator has verified your declaration, it counts as keeping prompts, and the default policy sends no traffic to a provider that may keep them.