Performance issue with latest app version

The app performance has been deteriorated with the latest version. The speech to text is unable to capture the interviewer's voice and convert that to a text, In today's interview I had to manually type in chat to get answers.
Also, please get back the LLama model, which was better than other models. The current model, the fast one gives limited answers without analyzing the question, and the sonet and other models give a lot of junk responses and super slow. LLama was the best one.