Title Latency comparison — hosted Vapi vs. on-prem (il-central-1) for calls originating in Israel --- Description Hi Vapi engineering, We’re evaluating whether to keep using the hosted Vapi SaaS backend or switch to the on-prem deployment we’re piloting in AWS il-central-1. Our callers are all in Israel and low round-trip latency is our top KPI. Could you share the expected end-to-end latency budget (human utterance → STT → LLM → TTS → audio back) for: 1. Hosted Vapi backend (your default EU/US region) when the SIP ingress is in Israel. 2. On-prem Vapi backend running entirely in il-central-1 with the same SIP ingress. A rough breakdown (network vs. processing) or any recent benchmark numbers would be very helpful in deciding which setup meets our sub-1500 ms target for conversational turn-taking. Thanks!

Use theon-prem Vapi deployment in AWS il-central-1, it meets your sub-1500 ms latency goal more reliably than the hosted EU/US setup, with total round-trip latency of 340–1,040 ms vs 500–1,300 ms.

I attempted to deploy on-premises, but it requires Nitro Enclave support. Unfortunately, AWS il-central-1 region doesn't support Nitro Enclave (I've created a separate ticket with additional details: https://discord.com/channels/1211482211119796234/1388180680533606572). @Kings_big💫 What would be the second-best deployment option while we wait for the Nitro Enclave dependency to be resolved?

Hey, I just replied to your another ticket. You can look at there for the answer related to Enclave support.

@User Which cloud provider and region do you use for hosting your EU infrastructure? Since AWS il-central-1 isn't an option, I'd like to deploy my custom TTS and STT services as close as possible to your setup to minimize latency for end users, while avoiding the complexity of on-premises deployment.

Hey, we have hosted our servers in the US region only. We are not deploying to EU yet.

1. Are there any plans for this in the near future? 2. Could you clarify what @Kings_big💫 meant by "hosted EU/US setup"? 3. What are the performance implications of having servers exclusively in the US? Specifically: - Do European users experience a full round-trip to US servers for each conversational exchange? - Or have you implemented a distributed architecture to handle real-time conversation components locally? 4. Would deploying on-premise infrastructure in Europe (such as Frankfurt, Germany) significantly reduce latency for EU users, and what would be the expected improvement?

2. Yes( US region )

Yes, we do have plans to host our VAPI servers in the EU, but unfortunately, it won't be in the next quarter; it might be next year. You won't notice any performance issues related to round trip time with servers hosted in the US—it's a negligible difference. We don't have a distributed architecture to handle real-time conversation components locally, as we are using service providers deployed in the US region. If you're experiencing high latency due to your own keys, deploying VAPI as an on-prem solution in Europe won't help. If you could provide a call ID with a timestamp where you're facing a latency issue, I can take a look and try to help you out.