Hi! I recently switched to gemini-3-flash for some testing. Before deploying to the cloud, I ran a few local tests and the latency looked pretty reasonable, with time-to-first-token around <1s. After deploying to the cloud, though, I’m seeing double that TTFT. What’s surprising is that my local tests were done on a pretty bad network, so I was expecting the cloud to perform better. For context, my LiveKit region is eu-central. I’m using the Gemini API (not Vertex), since Vertex seemed even slower in my tests. Is this kind of behavior expected?