hello guys.. I want to ask what llm model you are currently using for e.g. outbound voicebot? (where the fastest possible response is needed). In English it almost doesn't matter, because TTS are fast, but for smaller languages ​​like Czech, Slovak the latency is still sometimes 4 seconds which is too long in some cases. I am currently using Lama 4 Maverick 17b 128e Instruct, that is the best response I have found in small languages, but sometimes the problem repeats itself and I feel that it is not smart enough. For voice I use 11labs flash 2.5 and transcriber from Azure. Do you have any better ideas please? Thank you