Slow Latency (3000ms, triple what it should be)

Support Help
1

I’m consistently getting around 3000ms latency despite my agent being configured for 770-1100ms. How can I solve this? I’ve tried with Gemini 2.5 Flash and Gemini 3.0 Flash. I just tried one with Claude 4.5 Haiku as well and it wasn’t much better with 2403ms. I also tried GPT 4.1 Fast Tier and it was 2652ms. It doesn’t seem to be LLM specific.

If it helps, Workspace ID org_qOGEb7AZzoFqy48u

Call ID example:

call_2d8155894c27b6a37ecb3f5c5eb
call_5db5a24ccda664a27b8e8edb0bf
call_8070f738b1f171d3216a3c147d4
call_5b72f1d3118c51709376ce9d64e

2

For end-to-end latency exceeding 3 seconds when your estimated latency is 770-1100ms, follow these troubleshooting steps:(1)

1. Check Estimated Latency
Look for features marked with a turtle icon :turtle: in your configuration.(1) Check if your estimated latency is higher than 1.5s.(1) If high, disable some features to reduce latency.(1)

2. Monitor LLM Response Time
Check your current LLM latency and compare against the normal range of 500ms-900ms.(1) If consistently above 900ms, use fast tier, switch to an alternative LLM provider, or wait for your current provider to resolve performance issues.(1)

3. Check Status Page
Check the Status Page for any ongoing issues.(1) If there are any ongoing issues, wait for them to resolve.(1)

4. Consider Geographic Distance
International phone calls may introduce additional latency due to geographic distance.(1) Check if you’re making calls between different countries or continents.(1) Consider getting a local phone number in the same region as your users.(1) Use phone numbers geographically closer to your target audience for better performance.(1)

5. Turn-Taking Considerations
Extra end-to-end latency can sometimes be explained by interactions not being standard and turn-taking taking longer.(2) If the user says something brief while the system still assumes the user has not finished speaking yet (turn taking returns false), the agent waits, and that “pause” time is counted towards end-to-end latency.(2)

6. Contact Support
If the above steps don’t resolve your latency issues, locate your call ID and message the support-questions channel in the Discord community.(1) Include your call ID, steps you’ve already tried, and current latency measurements.(1)

3

I thought I’d hear more than an AI response that didn’t take into context what I said in my message.

4

Hello @cpaterson
I’ve checked your agent, and the latency seems to be coming from the TTS. I notice you are using Cartesia Sonic-2. I would advise using Sonic-3, and also try another provider like Elevenlabs it might give a better latency