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

### Answer GPT-5.6-Terra is listed as the default model for post call analysis, not as a recommended voice AI LLM in general [(1) ]() . The documentation states: "The model to use for post call analysis. Default to gpt-5.6-terra" , which applies specifically to the

post_call_analysis_model

field, not to the response engine LLM used for the voice conversation itself. I don't have information in the sources explaining why gpt-5.6-terra was chosen as this default, or any documentation recommending it as "the" voice AI LLM overall.

​ To learn more, see the sources I considered: [[1] Get Voice Agent](https://docs.retellai.com/api-references/get-agent) ​

​ To learn more, see the sources I considered: [[1] 2025 B2B Guide to AI Phone Calls & Voice Agents | Retell AI | Retell AI](https://www.retellai.com/blog/b2b-guide-to-ai-phone-calls) [[2] Check estimated latency](https://docs.retellai.com/reliability/check-estimated-latency) [[3] Check actual latency](https://docs.retellai.com/reliability/check-actual-latency) ​

### Answer I don't have a source that names a single "best" voice AI LLM. The sources only note that the LLM you choose determines how smart, fast, and relevant your agent sounds, with different models varying in latency, context retention, and ability to handle specialized vocabulary — platforms that let you plug in custom LLMs or fine-tune on your domain data give a performance edge [(1) ]() . Since latency is a major factor in voice AI performance, Retell AI's own low-latency architecture is designed to minimize end-to-end delay [(2) ]() , and you can inspect real per-call latency breakdowns (including per-LLM latency) via the Get Call API [(3) ]() . If you want, I can pull more detail on how to compare LLM latency for your specific setup.

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