### Answer Your latency of 1350-1850ms is above the ideal target of ~800ms but not critical—production voice agents typically aim for 800ms or lower to maintain conversational flow. [(1) ]() Anything beyond 800ms starts feeling unnatural to users. [(2) ]() ## To Decrease Latency *1. Check for features marked with 🐢 (turtle icon) * in your agent configuration and disable non-essential ones if your estimated latency exceeds 1.5s. [(3) ]() 2. Monitor your LLM response time — normal range is 500ms-900ms. If consistently above 900ms: - Use the Fast Tier option for OpenAI models - Switch to an alternative LLM provider 3. Consider geographic distance — international calls introduce additional latency. Use phone numbers geographically closer to your target audience. ## Regarding Token Count (2251-2491) High token counts typically correlate with longer LLM processing times, which directly impacts latency. To reduce tokens: - Shorten your system prompts - Use LLM States to split conversations into distinct stages with focused prompts - Reduce context window size where possible ## Will This Affect Calls? Yes—high latency leads to customer frustration, disrupted conversation flow, and lower trust in the AI system. [(2) ]() Users may talk over the agent, repeat themselves, or assume the call dropped. Check your actual latency metrics in Call History where you can see P90, Median, and Min values. [(4) ]()

​ To learn more, see the sources I considered: [[1] ](https://www.retellai.com/resources/sub-second-latency-voice-assistants-benchmarks) [[2] Latency](https://www.retellai.com/glossary/latency) [[3] Troubleshoot high latency](https://docs.retellai.com/reliability/troubleshoot-latency) [[4] Check actual latency](https://docs.retellai.com/reliability/check-actual-latency) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer Your P90 end-to-end latency of 4146ms exceeds the 3-second threshold that triggers troubleshooting. [(1) ]() ## Primary Issue: LLM Latency Your LLM latency (P90: 1578ms, Median: 1153ms) is the main contributor—normal range is 500ms-900ms. Since it's consistently above 900ms, you should: - Use the fast tier - Switch to an alternative LLM provider - Or wait for your current provider to resolve performance issues ## Other Components Look Acceptable - Transcription latency (P90: 130ms) — good - TTS latency (P90: 387ms) — reasonable - Knowledge base latency (P90: 267ms) — acceptable ## Additional Steps 1. Check for features marked with 🐢 in your configuration—if estimated latency exceeds 1.5s, disable some features 2. Check the [Status Page]() for ongoing issues 3. Consider geographic distance—international calls add latency. Use phone numbers geographically closer to your target audience ## If Issues Persist Locate your call ID and share it here with the steps you've already tried and your current latency measurements.

​ To learn more, see the sources I considered: [[1] Troubleshoot high latency](https://docs.retellai.com/reliability/troubleshoot-latency) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.