The machines that run the models sit behind NAT, so they dial out to this server. A request goes down that connection to whichever machine has room, and the reply comes straight back.
Shared or private
The shared pool
Sign in with an invitation and your token can use any machine in the pool. In return, other people can use yours.
Your own hardware
Keep a machine to yourself and give a token access to only that machine. You can still reach it from anywhere, but it's never lent to the pool.
Tokens processed
Input 998 M
Output 3.6 M
Counted over the last 90 days.
The pool
Machines lent to this server, counted by type. Which machine belongs to whom is never shown.
| Type | Online | Enrolled |
|---|---|---|
| DGX Spark | 1 | 1 |
| Mac Studio M2 Max | 0 | 1 |
| RTX 4080 | 0 | 1 |
| RX 6900 XT | 0 | 1 |
| RX 9070 XT | 0 | 1 |
| Strix Halo | 0 | 1 |
| Total | 1 | 6 |
API endpoint
Point any OpenAI-compatible client at this base URL. Signing in gets you a token to use as its API key.
https://ai.zygoon.pl/v1