When several providers serve the same model, choose which nodes your requests are routed to. Applies to all your inference requests.
Route to any online node that serves the model and that you have access to. The default — maximizes availability across the network.
Use one of your own nodes whenever it serves the requested model. Fall back to the rest of the network only when none of yours do.
Never route to other members. If none of your own nodes serve the model, the request fails instead of going elsewhere.
Tip: pair “Only my own nodes” with the Access settings to control who else can use the models you serve.