Hey, I am confused on the benefits of using our own Open AI key in vapi model provider vs using the vapi's default provider for evaluations. How does it looks on the rate limits side, and what is the benefit of using one over the other ?

1. lesser latency so faster response 2. more websocket connection 3. more tokens 4 .optmiized infra