I wouldn’t call it an air gap, but when you use a model on Bedrock or Vertex the inference runs on Amazon or Google’s servers respectively. The model provider gives them the weights, and is not otherwise in the loop for serving individual requests.
Not sure, why do you ask?
Because an air gap is a literal air gap, no networking, disconnected systems. It's not possible for models to function, with my understanding of it?
Oh they can, no problem. You just download (transfer data for) the weights and whatever you want running it.
You can install Gemma4 on ollama on your own computer and try it out! Lots of examples of people coding with no internet using qwen on flights etc.