The usual way to study Speaking is to go straight to the sample answer, listen, and note what was good. It sounds good at the time, and very little of it stays.
There is a far better way to use a model answer.
Do it in reverse
Before you listen to the model, give the answer yourself first.
Take a prompt: Describe a healthy habit you have recently developed.
Pause, open an AI assistant, and say:
I'm doing a speaking test. Could you help me? I'm going to tell you my answer and then you give me some feedback.
Then speak your answer into it. Naturally, a minute or two, without heavy preparation.
Why this does two jobs at once
First, you are actually practising speaking, rather than sitting and listening to someone else speak.
Second, and this is the valuable part: you now have a transcript of your own answer.
Then listen to the model answer. Then put your transcript next to it and compare.
The differences jump out: where they said something more economically, where they used a different tense from you, where you repeated a word without noticing.
Read the prompt properly too
That prompt contains the words recently developed. "Recently" is a signal: this is a question where the present perfect has a great deal of room to work.
Reading the question carefully before you speak is part of the practice.