What models are you using for voice AI bots and why?

Support Help
1

Hi everyone,

I’m curious what LLM models other teams are currently using for voice AI bots.

We’re actively evaluating model performance for production voice flows, especially around stability, latency, objection handling, natural conversation, and staying within the expected flow.

Would love to hear from the community:

What model are you using today for voice bots?
Why did you choose it?
Have you tested alternatives, and what were the biggest trade-offs?

Any practical insights, lessons learned, or recommendations would be super helpful.

Thanks!

2

We are using OpenAI GPT 4.1 for most of our agents. We’ve tried 5.5, but still not stable enough like 4.1, and in terms of cost, 4.1 is more affordable.

3

thanks for sharing, we are using 4.1 mini for better latency, have you faced any latency problems for voice bots?

and yeah - regarding new models, we’ve tried 5.4 but it was so unstable

4

There is latency, but still manageable. We are more focused on conversation quality.